List of Archived Posts
2026 Newsgroup Postings (07/11 - )
- Network History
- IBM Mainframe Assembler and REX
- Channel Throughput
- Future System, 3081, Amdahl
- Slow 3880 Controller
- CP67 and ASCII TTY
- CP67 and ASCII TTY
- SCI, FCS, HA/CMP
- IBM Mainframe Demise
- IBM Mainframe Demise
- IBM Mainframe HASP
- IBM Terminal Support
- IBM 370 CISC Chip
- Remember when discussions were worth reading?
- IBM Publication
- IBM Unbundling and HONE
- IBM 370/125
- NSF Supercomputer and Network
- NSF Supercomputer and Network
- NSF Supercomputer and Network
- ACP/TPF
- CMSBACK, ADSM, TSM
- NSFNET
- Tandem
- MVT Storage Management Problems
- MVS
- MVS
- HSDT
- HSDT
- HSDT
- SQL Server
- Punch Cards
- MIT CTSS/7094, MULTICS, UNIX, Virtual Machines
- Internet Lore
- Cloud Computing
- Cloud Computing
- VM/370
- Lynn Conway
- Reliability, Availability, Serviceability
- history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
- history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
- history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
- position-independent code (was Re: CMS, Self-hosting and the
- position-independent code (was Re: CMS, Self-hosting and the 6502)
- history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
- history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
- IBM Networking
- IBM 370 Virtual Memory
Network History
From: Lynn Wheeler <lynn@garlic.com>
Subject: Network History
Date: 11 Jul, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026b.html"#107 Network History
great change-over to TCP/IP (from IMPS and HOST protocol) was 1Jan1983
at the time there were about 100 IMP nodes with 255 connected HOSTS.
2nd half of 80s, I was on Chesson's XTP TAB. There were several gov
projects and so took XTP (as HSP) to (ISO chartered) ANSI X3S3.3 for
standardization (at time gov. was pushing for elimination TCP/IP and
everything move to GOSIP). At 1st they agreed ... but then they said
that ISO said no work on anything that didn't conform to OSI
model. XTP/HSP didn't because 1) supported internetworking layer
(which wasn't in OSI), 2) skip Transport/Networking interface, 3) went
directly to LAN/MAC interface which doesn't exist in OSI.
Joke at the time was ISO did standards that didn't even have to be
implementable while IETF (internet) required at least two
interoperable implementations for progression in standards process.
After leaving IBM in the 90s, (IETF RFC Editor) Postel let me help
with the periodically re-issued STD1. After NETSCAPE brought me into
do the stuff for interfacing to the payment networks, I did a talk on
why internet wasn't business critical dataprocessing (based on
documentation, software, processes I had do for financial
transactions), Postel sponsored my talk at USC/ISI.
OSI: The Internet That Wasn't. How TCP/IP eclipsed the Open Systems
Interconnection standards to become the global protocol for computer
networking
https://spectrum.ieee.org/osi-the-internet-that-wasnt
Meanwhile, IBM representatives, led by the company's capable director
of standards, Joseph De Blasi, masterfully steered the discussion,
keeping OSI's development in line with IBM's own business
interests. Computer scientist John Day, who designed protocols for the
ARPANET, was a key member of the U.S. delegation. In his 2008 book
Patterns in Network Architecture(Prentice Hall), Day recalled that IBM
representatives expertly intervened in disputes between delegates
"fighting over who would get a piece of the pie.... IBM played them
like a violin. It was truly magical to watch."
... snip ...
XTP/HSP posts
https://www.garlic.com/~lynn/subnetwork.html#xtphsp
web server support for payment transactions posts
https://www.garlic.com/~lynn/subnetwork.html#gateway
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
recent posts mentioning internet wasn't business critical
dataprocessing
https://www.garlic.com/~lynn/2025e.html#92 Internet
https://www.garlic.com/~lynn/2025e.html#79 IBM 8100, SNA, OSI, TCP/IP, Amadeus
https://www.garlic.com/~lynn/2025e.html#38 Amazon Explains How Its AWS Outage Took Down the Web
https://www.garlic.com/~lynn/2025e.html#35 Linux Clusters
https://www.garlic.com/~lynn/2025e.html#23 IBM Token-Ring
https://www.garlic.com/~lynn/2025d.html#111 ARPANET, NSFNET, Internet
https://www.garlic.com/~lynn/2025d.html#42 IBM OS/2 & M'soft
https://www.garlic.com/~lynn/2025b.html#97 Open Networking with OSI
https://www.garlic.com/~lynn/2025b.html#41 AIM, Apple, IBM, Motorola
https://www.garlic.com/~lynn/2025b.html#0 Financial Engineering
https://www.garlic.com/~lynn/2025.html#36 IBM ATM Protocol?
https://www.garlic.com/~lynn/2024g.html#80 The New Internet Thing
https://www.garlic.com/~lynn/2024g.html#71 Netscape Ecommerce
https://www.garlic.com/~lynn/2024g.html#16 ARPANET And Science Center Network
https://www.garlic.com/~lynn/2024e.html#43 Netscape
https://www.garlic.com/~lynn/2024d.html#97 Mainframe Integrity
https://www.garlic.com/~lynn/2024d.html#50 Architectural implications of locate mode I/O
https://www.garlic.com/~lynn/2024d.html#47 E-commerce
https://www.garlic.com/~lynn/2024c.html#92 TCP Joke
https://www.garlic.com/~lynn/2024c.html#62 HTTP over TCP
https://www.garlic.com/~lynn/2024b.html#73 Vintage IBM, RISC, Internet
https://www.garlic.com/~lynn/2024b.html#70 HSDT, HA/CMP, NSFNET, Internet
https://www.garlic.com/~lynn/2024b.html#33 Internet
https://www.garlic.com/~lynn/2024.html#71 IBM AIX
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Mainframe Assembler and REX
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Mainframe Assembler and REX
Date: 13 Jul, 2026
Blog: Facebook
Took two credit hr intro to fortran/computers. At the end of the
semester, I was hired to re-implement 1401 MPIO (709 tape->tape, 1401
reader/punch/printer front-end) in assembler for 360/30. Univ. was
getting 360/67 for tss/360 replacing 709/1401 and got 360/30
temporarily until availability of 360/67s. Univ. shutdown datacenter
on weekends and I got the whole place dedicated (although 48hrs w/o
sleep made Monday classes hard). I was given pile of hardware &
software manuals (learn assembler and hardware) and got to design and
implement my own monitor, device drivers, interrupt handlers, error
recovery, storage management, etc ... and within a few weeks had 2000
card 360/30 assembler program. Assembler option for two modes; 1)
stand-alone, IPLed with BPS loader and 2) OS/360 macros, GET/PUT,
OPEN/CLOSE, etc. Sometimes when I came in on weekend, production had
finished early and everything was dark/shutdown ... and 360/30 powerup
wouldn't work. Some manuals and experimenting, I found if I put all
the controllers in CE-mode, I could power on the 360/30, then power on
individual controllers and take them out of CE-mode.
The 360/67 arrived with-in a year of taking intro class and I was
hired fulltime responsible for OS/360 (TSS/360 not coming to
production). 709 use to do student fortran in less than second. 360/67
(as 360/65) os/360 (MFT9.5) took over a minute, I install HASP cutting
time in half. MFT11, I start redoing SYSGEN STAGE2 to carefully place
datasets and PDS members to optimize arm seek and (PDS directory)
multi-track search, cutting another 2/3rds to 12.9secs. OS/360 never
got better than 709 until I install Univ Waterloo WATFOR; on 360/65
WATFOR clocked at 20,000 statements/min (333/sec), student fortran
tended to run 30-60 statements.
Then IBM CSC came out to install (virtual machine) CP/67 (3rd after
CSC itself and MIT Lincoln Labs) and I mostly get to play with it
during my weekend 48hr window. I then spend a few months rewriting
pathlengths for running OS/360 in virtual machine. Bare machine test
ran 322secs ... initially 856secs (CP67 CPU 534secs). After a few
months I had CP67 CPU down from 534secs to 113secs. I then start
rewriting the dispatcher/scheduler , (dynamic adaptive resource
manager/default fair share scheduling policy), paging, adding ordered
seek queuing (from FIFO) and mutli-page transfer channel programs
(from FIFO and optimized for transfers/revolution, getting 2301 paging
drum from 70-80 4k transfers/sec to channel transfer peak of 270). Six
months after univ initial CP/67 install, CSC was giving one week class
in LA. I arrive on Sunday afternoon and asked to teach the class, it
turns out that the people that were going to teach it had resigned the
Friday before to join one of the 60s CSC CP67 commercial online
spin-offs.
CP/67 arrived with 1052&2741 terminal support and auto-terminal ident,
capable switching terminal type scanner type for each port. Univ. also
had TTY33&35 and I add ASCII support integrated with auto-terminal
type. I then want to have single dial-in number (hunt group) for all
terminals. Didn't quite work, IBM had hard-wired line speed ... so we
start a clone terminal controller. Build a IBM channel interface board
for Interdata/3 programmed to emulate IBM controller with the addition
for auto-baud. Then upgraded with Interdata/4 for channel interface
and cluster of Interdata/3s for port interfaces. Interdata (and later
Perkin-Elmer) sell them as clone controllers.
https://en.wikipedia.org/wiki/Interdata
https://en.wikipedia.org/wiki/Perkin-Elmer#Computer_Systems_Division
... and four of us are written up responsible for (some part of) IBM
clone controller business
Turn of century, toured datacenter that had descendant of our
Interdata box, that was handling nearly all card POS terminal dial-up
east of the Mississippi.
When I graduate, I join IBM CSC. A few years later, I transfer to SJR
on west coast and work with Jim Gray and Vera Watson on the original
SQL/relational RDBMS, System/R (morphs into both SQL/DS and DB2). I
also get to wander around datacenters in silicon valley, including
disk bldg14 engineering and bldg15 product test across street. They
were running pre-scheduled, 7x24, stand-alone testing and mentioned
they had tried MVS, but it had 15min MTBF (in that environment,
requiring manual re-IPL). I offer to rewrite I/O supervisor to make it
bullet proof and never fail, enabling any amount of on-demand,
concurrent testing, greatly improving productivity. I then write a I/O
Integrity research report and happen to mention MVS 15min MTBF,
bringing down the wrath of the MVS organization down on my head
(including blocking corporate award for the work)
Bldg15, product test gets the 1st engineering 3033 outside POK 3033
processor engineering. Product test was only taking a percent or two
of processing, so we scrounge up a 3830 controller and 3330 string for
our own private online service (also run 3270 coax under the street to
my office in bldg28). The 303x channel director was periodically still
hanging and required somebody to walk over and hit IMPL button. I then
find if I quickly hit all six channel addresses with CLRCH, it would
automagically re-IMPL. Note for 303x channel director, they took
158-engine with just the integrated channel microcode (and no 370
microcode). A 3031 was two 158-engines, one with just 370 microcode
and one with just integrated channel microcode. A 3032 was 168-3
reworked to use 303x channel director for external channels. A 3033
started out with 168 logic remapped to 20% faster chips.
More than decade after joining IBM, I wanted to show REX (before
renamed and released to customers) wasn't just another pretty
scripting language. I selected a large assembler dump analysis, IPCS
... to redo in REX with ten times the function and ten times the
performance, working half time over three months. I finished early and
wrote a automated library that looked for common failures. I had
assumed it would be released, replacing the assembler version ... but
for various reasons it wasn't (even though most internal datacenters
and PSRs were using it). I eventually got permission to give talks at
user group meetings on how it was done (and non-IBM versions started
appearing). Later the 3090 service processor (3092) planned to start
shipping it.
IBM CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
clone controller posts
https://www.garlic.com/~lynn/submain.html#360pcm
original SQL/Relational, System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
getting to play disk engineer posts
https://www.garlic.com/~lynn/subtopic.html#disk
REX DUMPRX posts
https://www.garlic.com/~lynn/submain.html#dumprx
A few recent posts mentioning bringing down the wrath of the MVS
organization on my head
https://www.garlic.com/~lynn/2026b.html#96 IBM CEOs and IBM Downfall
https://www.garlic.com/~lynn/2026b.html#95 Future System, 16-CPU, 3033, 3081, 3880, 3380
https://www.garlic.com/~lynn/2026b.html#70 IBM Online Systems
https://www.garlic.com/~lynn/2026b.html#57 IBM IPL and IMPL
https://www.garlic.com/~lynn/2026b.html#54 VM4341 Cluster Supercomputing and Departmental Distributed
https://www.garlic.com/~lynn/2026b.html#50 CSC, CP40/CMS, CP67/CMS, Boeing Computer Services
https://www.garlic.com/~lynn/2026b.html#16 IBM 16-CPU SMP
https://www.garlic.com/~lynn/2026b.html#4 IBM Virtual Machine
https://www.garlic.com/~lynn/2026b.html#2 IBM 370/195
https://www.garlic.com/~lynn/2026.html#82 IBM DASD, CKD, FBA
https://www.garlic.com/~lynn/2026.html#71 IBM 3033
https://www.garlic.com/~lynn/2026.html#68 Adirondack
https://www.garlic.com/~lynn/2026.html#60 IBM 135/145, 138/148, 4331/4341
https://www.garlic.com/~lynn/2026.html#59 IBM CP67 and VM370
https://www.garlic.com/~lynn/2026.html#50 Online Timesharing
https://www.garlic.com/~lynn/2026.html#34 IBM, NSC, HSDT, HA/CMP
https://www.garlic.com/~lynn/2026.html#33 IBM, NSC, HSDT, HA/CMP
https://www.garlic.com/~lynn/2026.html#28 360 Channel
https://www.garlic.com/~lynn/2025e.html#106 CP67, VM370, MVT, VS2/SVS, VS2/MVS, MVS/XA
https://www.garlic.com/~lynn/2025e.html#89 Re-impl Controllers
https://www.garlic.com/~lynn/2025e.html#75 Interactive response
https://www.garlic.com/~lynn/2025e.html#62 IBM Mainframe Projects
https://www.garlic.com/~lynn/2025e.html#47 IBM 360/85
https://www.garlic.com/~lynn/2025e.html#14 IBM DASD, CKD and FBA
https://www.garlic.com/~lynn/2025e.html#10 IBM Interactive Response
https://www.garlic.com/~lynn/2025d.html#107 Rapid Response
https://www.garlic.com/~lynn/2025d.html#98 IBM Supercomputer
https://www.garlic.com/~lynn/2025d.html#45 Some VM370 History
https://www.garlic.com/~lynn/2025d.html#35 IBM Internal Apps, Retain, HONE, CCDN, ITPS, Network
https://www.garlic.com/~lynn/2025d.html#34 IBM Internal Apps, Retain, HONE, CCDN, ITPS, Network
https://www.garlic.com/~lynn/2025c.html#92 FCS, ESCON, FICON
https://www.garlic.com/~lynn/2025c.html#62 IBM Future System And Follow-on Mainframes
https://www.garlic.com/~lynn/2025c.html#53 IBM 3270 Terminals
https://www.garlic.com/~lynn/2025c.html#47 IBM 3270 Terminals
https://www.garlic.com/~lynn/2025c.html#42 SNA & TCP/IP
https://www.garlic.com/~lynn/2025c.html#29 360 Card Boot
https://www.garlic.com/~lynn/2025c.html#2 Interactive Response
https://www.garlic.com/~lynn/2025.html#113 2301 Fixed-Head Drum
https://www.garlic.com/~lynn/2025.html#86 Big Iron Throughput
https://www.garlic.com/~lynn/2025.html#71 VM370/CMS, VMFPLC
https://www.garlic.com/~lynn/2025.html#61 old pharts, Multics vs Unix
https://www.garlic.com/~lynn/2025.html#59 Multics vs Unix
--
virtualization experience starting Jan1968, online at home since Mar1970
Channel Throughput
From: Lynn Wheeler <lynn@garlic.com>
Subject: Channel Throughput
Date: 13 Jul, 2026
Blog: Facebook
3272 channel attach controller (for 3277 terminals) had .086 hardware
response. for 3278 they move a lot of the electronics to the 3274 that
drove up the coax protocol chatter and hardware response went to .3-.5
secs (depending on how much data involved). 3274/3278 time-frame there
were studies about quarter second response improved
productivity. Letters to 3278 product administrator got response that
3278 wasn't intended for interactive computing, but "data entry".
after joining IBM CSC, one of my hobbies was enhanced production
operating systems for internal datacenters (online branch
sales&marketing HONE was one of my 1st and long time customers). At
the time 3278 introduced I had several customers getting .11 trivial
interactive system response (w/3277, aggregate for humans as .086+.11
or .196secs ... meeting .25sec requirement). MVS users rarely noticed
was because very few MVS systems even met one sec system response).
1980, STL was bursting at the seams and they were moving 300 from IMS
group to offsite bldg. They had tried "remote" (telco/SNA) 3270 and
found human factors were totally unacceptable. I get con'ed into doing
channel-extender support ... allowing channel attached controllers to
be placed at offsite bldg ... resulting in no noticeable human factors
difference (with inside STL).
There was unintended consequences ... STL had distributed 3270
controllers across all 168 channels with 3330s. 3270 controllers had
high channel busy interfering with 3330 I/O. The channel-extender had
significantly lower 168 channel busy ... improving 3330 I/O and
increasing system throughput 10-15%. There was consideration on moving
*ALL* 3270 controllers (even inside STL) to channel-extender.
In 1988, IBM branch asks if I can help LLNL (national lab) with
standardization of some serial stuff they were working with ... which
quickly becomes fibre-channel standard ("FCS", including some stuff I
had done in 1980, initially 1gbit transfer, full-duplex, aggregate
200mbyte/sec). Then IBM POK mainframe releases some serial (when it
was already obsolete) as ESCON, initially 10mbyte/sec. Then some POK
engineers become involved with "FCS" and define a heavy-weight
protocol that drastically cuts native throughput, eventually ships as
FICON.
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
channel-extender posts
https://www.garlic.com/~lynn/submisc.html#channel.extender
FCS and/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
--
virtualization experience starting Jan1968, online at home since Mar1970
Future System, 3081, Amdahl
From: Lynn Wheeler <lynn@garlic.com>
Subject: Future System, 3081, Amdahl
Date: 16 Jul, 2026
Blog: Facebook
Amdahl won the battle to make ACS, 360 compatible. When ACS/360 was
killed (folklore: executives were afraid it would advance state of art
too fast and IBM would loose control of the market), Amdahl leaves
IBM. Following lists some ACS/360 features that don't show up until
the 90s with ES/9000.
https://mark.people.clemson.edu/acs_end.html
Future System effort 1st half of 70s was completely different than 370
and planned to replace 370. Internal politics was killing off 370
efforts ... the lack of new 370s during FS is credited with giving
clone 370 makers (including Amdahl) their market foothold. When FS
implodes, there is mad rush to get stuff back into product pipelines,
including kicking off quick&dirty 303x and 3081 efforts in parallel.
https://en.wikipedia.org/wiki/IBM_Future_Systems_project
https://people.computing.clemson.edu/~mark/fs.html
... and from "Computer Wars: The Post-IBM World"
https://www.amazon.com/Computer-Wars-The-Post-IBM-World/dp/1587981394/
... and perhaps most damaging, the old culture under Watson Snr and Jr
of free and vigorous debate was replaced with *SYNCOPHANCY* and *MAKE
NO WAVES* under Opel and Akers. It's claimed that thereafter, IBM
lived in the shadow of defeat ... But because of the heavy investment
of face by the top management, F/S took years to kill, although its
wrong headedness was obvious from the very outset. "For the first
time, during F/S, outspoken criticism became politically dangerous,"
recalls a former top executive
... snip ...
ASDD Products Architecture
http://www.jfsowa.com/computer/memo125.htm
In 1973, I had attended a presentation on the design of the VANDERBILT
hardware. After showing a diagram of the CPU frame, which contained
sixteen circuit boards, the presenter proudly said "Each of these
boards contains as much circuitry as a System/370, Model 168." At that
point, I asked the obvious question: "But the expected performance is
only about three times faster than a Model 168. Why don't you just
build sixteen 168s?" Instead of answering, he just glared at me. But
when the FS project was canceled in 1975, IBM did just that. According
to the IBM archives, "Whereas the Model 168 had required 40 months to
evolve from development to initial shipment, the 3033 was shipped to
its first customer after only 28 months in development." IBM achieved
that feat by remapping the Model 168 design into the circuitry
intended for VANDERBILT and making a number of tweaks in the design to
improve performance. Of course, IBM could have delivered a machine
with similar or better performance in 1975 instead of 1977, if they
hadn't killed all the System/370 design projects to avoid competition
with the FS fantasy.
... snip ...
When FS imploded, I got asked to help with a 16-CPU 370 and we con the
3033 processor engineers to help in their spare time (lot more
interesting than remapping 168 logic to 20% faster chips). Everybody
thought it was great until somebody told the head of POK that it could
be decades before POK's favorite son operating system ("MVS") had
(effective) 16-cpu support (POK doesn't ship 16-CPU system until after
turn of the century). Head of POK then tells some of us to never visit
POK again and directed 3033 processor engineers, heads down and no
distractions (once the 3033 is out the door, they start on
trout/3090).
Initial 3081D 2-CPU aggregate MIPS was less than Amdahl single
processor system. IBM doubled the processor cache sizes for the 3081K
2-CPU, bringing aggregate MIPS up to about same as Amdahl
1-CPU. However MVS docs had MVS 2-CPU multiprocessor support only got
1.2-1.5 times throughput of 1-CPU ... so MVS 2-CPU 3081K throughput
was only .6-.75 of Amdahl 1-CPU MVS (despite having approx. same
aggregate MIPS).
also
http://www.jfsowa.com/computer/memo125.htm
The 370 emulator minus the FS microcode was eventually sold in 1980 as
as the IBM 3081. The ratio of the amount of circuitry in the 3081 to
its performance was significantly worse than other IBM systems of the
time; its price/performance ratio wasn't quite so bad because IBM had
to cut the price to be competitive. The major competition at the time
was from Amdahl Systems -- a company founded by Gene Amdahl, who left
IBM shortly before the FS project began, when his plans for the
Advanced Computer System (ACS) were killed. The Amdahl machine was
indeed superior to the 3081 in price/performance and spectaculary
superior in terms of performance compared to the amount of circuitry.]
... snip ...
... and claims that the enormous number circuits required the use of
TCMs in order to package the circuits in a reasonably sized space.
https://en.wikipedia.org/wiki/History_of_IBM#1980
not long after being told to never visit POK again, I transfer out to
SJR on west coast. I worked with Jim Gray and Vera Watson on original
SQL/relational, System/R (all work done on VM370). I also get to visit
silicon valley datacenters including disk bldg14 (engineering) and
bldg15 (product test) across the street. The were doing prescheduled,
7x24, stand-alone testing and said that they had recently tried MVS,
but it had 15min MTBF (requiring manual re-ipl in that environment). I
offer to rewrite I/O supervisor to make it bullet proof and never
fail, allowing any amount of on-demand, concurrent testing, greatly
improving productivity.
Then bldg15 get engineering 3033 (1st outside POK processor
engineering). Testing was only taking percent or two of CPU, so we
scrounge up a 3830 and 3330 string and setup our own private online
service. Then summer of 1978, get engineering 4341 (year before
announce). Branch office hears about it and Jan1979 cons me into doing
benchmark for national lab that was looking at getting 70 for compute
farm (sort of leading edge of the coming cluster supercomputing
tsunami).
Also find that small cluster of five 4341s had higher aggregate
throughput, much less expensive, much less power & cooling, and
smaller footprint than 3033. Also BofA had System/R joint study and
orders 60 VM4341s for distributed operation.
future system posts
https://www.garlic.com/~lynn/submain.html#futuresys
SMP, tightly-coupled, shared memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
original SQL/Relational RDBMS posts
https://www.garlic.com/~lynn/submain.html#systemr
getting to play disk engineer posts
https://www.garlic.com/~lynn/subtopic.html#disk
some recent posts mentioning Amdahl and cancel ACS/360
https://www.garlic.com/~lynn/2026b.html#100 Amdahl
https://www.garlic.com/~lynn/2026b.html#95 Future System, 16-CPU, 3033, 3081, 3880, 3380
https://www.garlic.com/~lynn/2026b.html#93 Hercules IBM Emulator
https://www.garlic.com/~lynn/2026b.html#73 IBM System/360
https://www.garlic.com/~lynn/2026b.html#68 IBM Career, Promotions, Raises
https://www.garlic.com/~lynn/2026b.html#51 Mainframe, Virtualization, Cloud
https://www.garlic.com/~lynn/2026b.html#18 IBM 16-CPU SMP
https://www.garlic.com/~lynn/2026b.html#2 IBM 370/195
https://www.garlic.com/~lynn/2026.html#71 IBM 3033
https://www.garlic.com/~lynn/2026.html#25 Amdahl Computers
https://www.garlic.com/~lynn/2026.html#24 IBM 360, Future System
https://www.garlic.com/~lynn/2026.html#15 IBM 360s, Unbundling, 370s, Future System
https://www.garlic.com/~lynn/2026.html#11 4341, cluster supercomputing, distributed computing
--
virtualization experience starting Jan1968, online at home since Mar1970
Slow 3880 Controller
From: Lynn Wheeler <lynn@garlic.com>
Subject: Slow 3880 Controller
Date: 18 Jul, 2026
Blog: Facebook
Note once the 3033 was out the door, the processor engineers start on
trout/3090. They configure memory, processor, channels, I/O to meet
specific throughput goals. However, in San Jose, the administration
dictated that the 3880 disk controller would use a (really slow,
resulting in significant channel busy) vertical CISC processor (much
slower than the previous processor in 3830 controller). The 3090
people thought the 3880 would be the same as 3830, but supporting 3380
3mbyte/sec transfer, when they found out how bad the 3880 controllers'
channel busy was, they realized they would have to significantly
increase the number of I/O channels in order to achieve system
throughput targets ... which required an additional TCM (3090 said
semi-facetiously they would bill the 3880 organization for the
increase in 3090 manufacturing cost). Marketing eventually comes up
with what a fantastic I/O machine the 3090 is with big increase in
number of channels.
2nd half 70s, I transfer out to SJR on west coast. I worked with Jim
Gray and Vera Watson on original SQL/relational, System/R (all work
done on VM370). I also get to visit silicon valley datacenters
including disk bldg14 (engineering) and bldg15 (product test) across
the street. The were doing prescheduled, 7x24, stand-alone testing and
said that they had recently tried MVS, but it had 15min MTBF
(requiring manual re-ipl in that environment). I offer to rewrite I/O
supervisor to make it bullet proof and never fail, allowing any amount
of on-demand, concurrent testing, greatly improving productivity. I
write a I/O Reliability Enhancement research report and happen to mention the MVS
15min MTBF, bringing down the wrath of the MVS organization on my
head. Then bldg15 get engineering 3033 (1st outside POK processor
engineering). Testing was only taking percent or two of CPU, so we
scrounge up a 3830 and 3330 string and setup our own private online
service.
One monday morning, I come in and have irate phone calls from bldg15
asking what did I do to the 3033 system over the weekend
... interactive CMS response had gone over a second (my internal
sjr/vm systems had .11 interactive system response). I said I did
nothing, "what did they do". They said nothing, eventually it turns
out somebody had replaced the 3830 with a 3880. Had to regress to the
3830 ... had at least six months before 3880 customer release and had
to come up with a few hacks to try and obfuscate some of the worst
side effect consequences of the slow 3880 processor.
"calypso" 3880 was to connect 3mbyte/sec 3380 to 1.5mbyte channels as
well as eckd channel commands.
engineering 3033 was 3880 connecting with 3330 disks. 3880 had done
acceptance with vanilla single thread/drive mvs test. mvs has
something 10k instruction path for channel redrive (from interrupt to
start nxt queued channel program) ... to help mask slow 3880, they
present ending interrupt early and figured to do controller cleanup
overlapped with operating system processing. single drive thread/drive
mvs didn't have queued channel programs so it was more than 10k to nxt
system i/o.
my sjr/vm had something like 500 instruction, interrupt to channel
redrive (and bldg15 heavy use could have multiple queued channel
programs waiting). Channel redrive then would immediately result in
CC=1,SM+BUSY (control unit busy), have to requeue the channel program
and wait for CUE (to retry channel redrive).
trivia: early last decade, I was asked to track down executive
decision to add virtual memory to all 370s, and found staff to the
executive. Basically MVT storage management was so bad, that REGION
sizes had to be specified four times larger than used. As a result
typical 1mbyte 370/165 only ran four concurrent regions, insufficient
to keep system busy and justified. going to MVT in 16mbyte virtual
address space, allowed number of regions to be increased by factor of
four (capped at 15 because of 4bit storage protect keys) with little
or not paging (sort of like running MVT in CP67 16mbyte virtual
machine).
I periodically dropped in Ludlow doing initial VS2/SVS on 360/67
(pending engineering 370 w/virtual memory). Little bit of code to
build virtual memory tables and some simple paging. Biggest was
channel programs passed to EXCP/SVC0 had virtual addresses (and
channels require real addresses). He borrows CP67 CCWTRANS (that
creates channel program copies with real addresses) for crafting into
EXCP.
So part of 370/XA was architecture to address the excess latency that
it took for MVS channel redrive.
Also 1988, branch office asks me if I could help LLNL (national lab)
with standardizing some serial stuff they had been working with
... which quickly becomes fibre-channel standard ("FCS", initial 1gbit
transfer, full-duplex, aggregate 200mbytes/sec, including some stuff I
had done in 1980). POK eventually releases their serial stuff as ESCON
(when it is already obsolete), initially 10mbyte/sec transfer. Then
some POK engineers become involved with "FCS" and define a
heavy-weight protocol that drastically cuts native throughput,
eventually ships as FICON.
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
getting to play disk engineer in bldgs 14&15 posts
https://www.garlic.com/~lynn/subtopic.html#disk
channel-extender posts
https://www.garlic.com/~lynn/submisc.html#channel.extender
FCS and/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
--
virtualization experience starting Jan1968, online at home since Mar1970
CP67 and ASCII TTY
From: Lynn Wheeler <lynn@garlic.com>
Subject: CP67 and ASCII TTY
Date: 21 Jul, 2026
Blog: Facebook
Took two credit hr intro to fortran/computers. At the end of the
semester, I was hired to re-implement 1401 MPIO (709 tape->tape,
1401 reader/punch/printer front-end) in assembler for
360/30. Univ. was getting 360/67 for tss/360 replacing 709/1401 and
got 360/30 temporarily until availability of 360/67s. Univ. shutdown
datacenter on weekends and I got the whole place dedicated (although
48hrs w/o sleep made Monday classes hard). I was given pile of
hardware & software manuals (learn assembler and hardware) and got
to design and implement my own monitor, device drivers, interrupt
handlers, error recovery, storage management, etc ... and within a few
weeks had 2000 card 360/30 assembler program. Assembler option for two
modes; 1) stand-alone, IPLed with BPS loader and 2) OS/360 macros,
GET/PUT, OPEN/CLOSE, etc. Sometimes when I came in on weekend,
production had finished early and everything was dark/shutdown ... and
360/30 powerup wouldn't work. Some manuals and experimenting, I found
if I put all the controllers in CE-mode, I could power on the 360/30,
then power on individual controllers and take them out of CE-mode.
The 360/67 arrived with-in a year of taking intro class and I was
hired fulltime responsible for OS/360 (TSS/360 not coming to
production). 709 use to do student fortran in less than second. 360/67
(as 360/65) os/360 (MFT9.5) took over a minute, I install HASP cutting
time in half. MFT11, I start redoing SYSGEN STAGE2 to carefully place
datasets and PDS members to optimize arm seek and (PDS directory)
multi-track search, cutting another 2/3rds to 12.9secs. OS/360 never
got better than 709 until I install Univ Waterloo WATFOR; on 360/65
WATFOR clocked at 20,000 statements/min (333/sec), student fortran
tended to run 30-60 statements.
Then IBM CSC came out to install (virtual machine) CP/67 (3rd
after CSC itself and MIT Lincoln Labs) and I mostly get to play with
it during my weekend 48hr window. I then spend a few months rewriting
pathlengths for running OS/360 in virtual machine. Bare machine
test ran 322secs ... initially 856secs (CP67 CPU 534secs). After a few
months I had CP67 CPU down from 534secs to 113secs. I then start
rewriting the dispatcher/scheduler , (dynamic adaptive resource
manager/default fair share scheduling policy), paging, adding ordered
seek queuing (from FIFO) and mutli-page transfer channel programs
(from FIFO and optimized for transfers/revolution, getting 2301 paging
drum from 70-80 4k transfers/sec to channel transfer peak of 270). Six
months after univ initial CP/67 install, CSC was giving one week class
in LA. I arrive on Sunday afternoon and asked to teach the class, it
turns out that the people that were going to teach it had resigned the
Friday before to join one of the 60s CSC CP67 commercial online
spin-offs.
CP/67 arrived with 1052&2741 terminal support and auto-terminal
ident, capable switching terminal type scanner type for each
port. Univ. also had TTY33&35 (paper tape) and I add TTY ASCII
support integrated with auto-terminal type (ASCII port scanner had
arrived in a Heathkit box; TTY required much more finger strength than
IBM selectric typeball-based 1052&2741). I then want to have
single dial-in number (hunt group) for all terminals. Didn't quite
work, IBM had hard-wired port line speed ... so we start a clone
terminal controller. Build a IBM channel interface board for
Interdata/3 programmed to emulate IBM controller with the addition for
auto-baud. Then upgraded with Interdata/4 for channel interface and
cluster of Interdata/3s for port interfaces. Interdata (and later
Perkin-Elmer) sell them as clone controllers.
https://en.wikipedia.org/wiki/Interdata
https://en.wikipedia.org/wiki/Perkin-Elmer#Computer_Systems_Division
... and four of us are written up responsible for (some part of) IBM
clone controller business
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
clone controller posts
https://www.garlic.com/~lynn/submain.html#360pcm
some recent posts mentioning 709, 1401 MPIO, Fortran, 360/67 and WATFOR
https://www.garlic.com/~lynn/2026c.html#1 IBM Mainframe Assembler and REX
https://www.garlic.com/~lynn/2026b.html#88 Fortran
https://www.garlic.com/~lynn/2026b.html#78 Early CSC (& CP67/CMS)
https://www.garlic.com/~lynn/2026b.html#73 IBM System/360
https://www.garlic.com/~lynn/2026b.html#72 IBM 2321 Datacell, CICS
https://www.garlic.com/~lynn/2026b.html#70 IBM Online Systems
https://www.garlic.com/~lynn/2026b.html#14 Bad Responsee
https://www.garlic.com/~lynn/2026b.html#4 IBM Virtual Machine
https://www.garlic.com/~lynn/2026.html#98 IBM 360&370 Experience
https://www.garlic.com/~lynn/2026.html#91 CP/67 and VM/370
https://www.garlic.com/~lynn/2026.html#83 Touch Type, Typewriters, Terminals
https://www.garlic.com/~lynn/2026.html#82 IBM DASD, CKD, FBA
https://www.garlic.com/~lynn/2026.html#81 IBM 60s-90s
https://www.garlic.com/~lynn/2026.html#78 IBM OS Debugging
https://www.garlic.com/~lynn/2026.html#67 Early Mainframe work
https://www.garlic.com/~lynn/2026.html#59 IBM CP67 and VM370
https://www.garlic.com/~lynn/2026.html#46 UofM MTS and IBM CP67
https://www.garlic.com/~lynn/2026.html#28 360 Channel
https://www.garlic.com/~lynn/2026.html#24 IBM 360, Future System
--
virtualization experience starting Jan1968, online at home since Mar1970
CP67 and ASCII TTY
From: Lynn Wheeler <lynn@garlic.com>
Subject: CP67 and ASCII TTY
Date: 23 Jul, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#5 CP67 and ASCII TTY
When I graduate, I join IBM Science Center, and get home (134.5 baud)
2741, which was replaced by a 300baud CDI Miniterm (very similar to TI
silent 700), which was quickly replaced with early (1200 baud) IBM
glass teletype, TOPAZ/3101. Then got ROMS from the plant in Japan that
supported "block mode" (3270 like operation).
https://en.wikipedia.org/wiki/IBM_3101
VM370 implemented virtual 3270 and "passthrough virtual machine
(PVM)" used interface to simulate 3270s over the internal network
... which also included gateway to the internal CCDN terminal
network. Then for the travel/home terminal program got special
encrypting 2400 baud modems (that could simulate industry standard
modem).
Then there was IBM/PC PCTERM program for VM370 home terminal service
that used PVM for host 3270 emulation and a IBM/PC app that simulated
3270 terminal. The host/PC (PCTERMs) maintained synchronized caches of
recently transmitted data,, could specify changes to the current
screen display (actual data, or pieces of data from cache), reducing
amount of data transmitted (inbound and outbound). For actual data
transmitted, it was further reduced using compression.
Also, late 70s, the (CMS 3270 app) VMSG (used by PROFS for the email
client) author had done (CMS) PARASITE/STORY (predating IBM/PC HLLAPI)
... interfaced to the VM370 virtual 3270 support. It could do
programmed 3270s on the same machine or interface to PVM. A STORY in
this archived post to log into the internal IBM "RETAIN" and download
(customer) support information.
https://web.archive.org/web/20260305093053/https://www.garlic.com/~lynn/2001k.html#36
a few PCTERM posts
https://www.garlic.com/~lynn/2025b.html#66 IBM 3101 Glass Teletype and "Block Mode"
https://www.garlic.com/~lynn/2025.html#75 IBM Mainframe Terminals
https://www.garlic.com/~lynn/2024f.html#43 IBM/PC
https://www.garlic.com/~lynn/2021h.html#33 IBM/PC 12Aug1981
https://www.garlic.com/~lynn/2016b.html#101 You count as an old-timer if (was Re: Origin of the phrase "XYZZY")
https://www.garlic.com/~lynn/2014j.html#25 another question about TSO edit command
https://www.garlic.com/~lynn/2014i.html#11 The Tragedy of Rapid Evolution?
https://www.garlic.com/~lynn/2014h.html#71 The Tragedy of Rapid Evolution?
https://www.garlic.com/~lynn/2014e.html#49 Before the Internet: The golden age of online service
https://www.garlic.com/~lynn/2011d.html#6 If IBM Hadn't Bet the Company
https://www.garlic.com/~lynn/2008n.html#51 Baudot code direct to computers?
https://www.garlic.com/~lynn/2007o.html#66 The use of "script" for program
https://www.garlic.com/~lynn/2006y.html#0 Why so little parallelism?
--
virtualization experience starting Jan1968, online at home since Mar1970
SCI, FCS, HA/CMP
From: Lynn Wheeler <lynn@garlic.com>
Subject: SCI, FCS, HA/CMP
Date: 23 Jul, 2026
Blog: Facebook
1988, branch office asks if I could help SLAC/Gustavson with SCI
standard (various uses including shared memory multiprocessor, up to
64 cache consistency, some efforts Data General and Sequent 64 4-i486
boards (256 processors), Convex 64 2-HP-snake boards (128 processor),
SGI 64 4-MIPS boards, etc),
https://en.wikipedia.org/wiki/Scalable_Coherent_Interface
https://www.scizzl.com/
... same time asked about helping LLNL (national lab) with
fibre-channel standard ("FCS", initial 1gbit transfer, full-duplex,
aggregate 200mbyte/sec)
https://en.wikipedia.org/wiki/Fibre_Channel
Then IBM announces their fiber stuff as ESCON (when it was aleady
obsolete), initially 10mbyte/sec. Then IBM engineers become involved
with FCS and define protocol that significantly cuts native
throughput, that is announced as FICON
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
... also 1988, Nick Donofrio approves HA/6000, initially for NYTimes
to move their newspaper system (ATEX) off DEC VAXCluster to RS/6000, I
rename it HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing (technical/scientific) cluster scale-up with
national labs (LANL, LLNL, NCAR/UCAR, etc) and commercial cluster
scale-up with RDBMS vendors (Oracle, Sybase, Ingres, Informix) with
VAXCluster support in same source base with UNIX. Also migrate LLNL
UNICOS "LINCS"
https://pdfs.semanticscholar.org/f5fc/36f760a62a39e1b780b556d5817841d40fac.pdf
and NCAR "Mesa Archive" spin-off, to HA/CMP.
Early Jan92, there was meeting with Oracle CEO and IBM/AWD executive
Hester tells Ellison that we would have 16-system clusters by mid92
and 128-system clusters by ye92. Mid-jan92, I update FSD on HA/CMP
work with national labs and FSD decides to go with HA/CMP for federal
supercomputers. By end of Jan, we are told that cluster scale-up is
being transferred to Kingston for announce as IBM Supercomputer (SP,
technical/scientific *ONLY*) and we aren't allowed to work with
anything that has more than four systems (we leave IBM a few months
later). A couple weeks later, 17feb1992, Computerworld news ... IBM
establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
Some speculation that it would have eaten the mainframe in the
commercial market. 1993 industry benchmarks (number of program
iterations compared to the industry MIPS/BIPS reference platform):
ES/9000-982 : 8CPU 408MIPS, (51MIPS/CPU)
RS6000/990 (RIOS chipset) : (1-CPU) 126MIPS, 16-systems: 2BIPS,
... 128-systems: 16BIPS
After leaving IBM, I do some consulting for Steve Chen (was
responsible for YMP at Cray), was then CTO at Sequent.
https://en.wikipedia.org/wiki/Sequent_Computer_Systems
SCI Exemplar
https://en.wikipedia.org/wiki/Sequent_Computer_Systems#NUMA
before IBM buys Sequent and shuts it down.
https://en.wikipedia.org/wiki/Sequent_Computer_Systems#IBM_purchase_and_disappearance
FCS and/or FICON Posts
https://www.garlic.com/~lynn/submisc.html#ficon
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
SCI, scalable.coherent.interface, and Sequent posts
https://www.garlic.com/~lynn/2026b.html#23 IBM CSC, IBM Unbundle, IBM HONE, IBM System/R, SCI, FCS, IBM HA/CMP
https://www.garlic.com/~lynn/2026.html#69 IBM System/R, SCI, FCS, IBM HA/CMP
https://www.garlic.com/~lynn/2026.html#66 IBM CSC, IBM Unbundle, IBM HONE, IBM System/R, SCI, FCS, IBM HA/CMP
https://www.garlic.com/~lynn/2026.html#54 IBM Supercomputer
https://www.garlic.com/~lynn/2025c.html#24 IBM AIX
https://www.garlic.com/~lynn/2025c.html#16 Cluster Supercomputing
https://www.garlic.com/~lynn/2024g.html#85 IBM S/38
https://www.garlic.com/~lynn/2024e.html#90 Mainframe Processor and I/O
https://www.garlic.com/~lynn/2024.html#54 RS/6000 Mainframe
https://www.garlic.com/~lynn/2024.html#37 RS/6000 Mainframe
https://www.garlic.com/~lynn/2023g.html#106 Shared Memory Feature
https://www.garlic.com/~lynn/2023g.html#22 Vintage Cray
https://www.garlic.com/~lynn/2023g.html#16 370/125 VM/370
https://www.garlic.com/~lynn/2023e.html#78 microcomputers, minicomputers, mainframes, supercomputers
https://www.garlic.com/~lynn/2023.html#42 IBM AIX
https://www.garlic.com/~lynn/2022g.html#91 Stanford SLAC (and BAYBUNCH)
https://www.garlic.com/~lynn/2022f.html#29 IBM Power: The Servers that Apple Should Have Created
https://www.garlic.com/~lynn/2022.html#118 GM C4 and IBM HA/CMP
https://www.garlic.com/~lynn/2022.html#95 Latency and Throughput
https://www.garlic.com/~lynn/2022.html#23 Target Marketing
https://www.garlic.com/~lynn/2021i.html#16 A brief overview of IBM's new 7 nm Telum mainframe CPU
https://www.garlic.com/~lynn/2021h.html#45 OoO S/360 descendants
https://www.garlic.com/~lynn/2021b.html#64 Early Computer Use
https://www.garlic.com/~lynn/2021b.html#44 HA/CMP Marketing
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Mainframe Demise
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Mainframe Demise
Date: 31 Jul, 2026
Blog: Facebook
Late 80s, senior gpd/adstar engineer gets a talk scheduled at
internal, annual, world-wide communication group conference,
supposedly on 3174 performance, but opens the talk with statement that
the communication group was going to be responsible for the demise of
the disk division. the disk division was seeing drop in disk sales
with data fleeing mainframes to more distributed computing friendly
platforms. the disk division had come up with a number of solutions,
but they were all being vetoed by the communication group with their
corporate responsibility for everything that crossed datacenter
walls. The senior gpd/adstar software executive had partial
countermeasure with investments in distributed computing startups that
would use IBM disks (and would periodically ask us to stop in on his
investments to offer help).
Articles started appearing about the demise of
mainframes. Communication group stanglehold on IBM datacenters wasn't
just disks and 1992 IBM has one of the largest losses in the history
of US companies. IBM was being reorganized into the 13 "baby blues" in
preparation for breaking up the company (take-off on the "baby bells"
breakup a decade earlier)
https://web.archive.org/web/20101120231857/http://www.time.com/time/magazine/article/0,9171,977353,00.html
https://content.time.com/time/subscriber/article/0,33009,977353-1,00.html
we had already left IBM but get a call from the bowels of Armonk
asking if we could help with the breakup of the company. Before we get
started, the board brings in the former president of Amex as CEO to
try and save the company, who (somewhat) reverses the breakup and uses
some of the same techniques used at RJR (gone 404, but lives on at
wayback)
https://web.archive.org/web/20181019074906/http://www.ibmemployee.com/RetirementHeist.shtml
Didn't save the disk division ... there hasn't been any IBM CKD DASD
manufactured for decades, all being simulated on industry standard
fixed-block devices.
Demise of disk division posts
https://www.garlic.com/~lynn/subnetwork.html#terminal
IBM downturn/downfall/breakup posts
https://www.garlic.com/~lynn/submisc.html#ibmdownfall
pension posts
https://www.garlic.com/~lynn/submisc.html#pension
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Mainframe Demise
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Mainframe Demise
Date: 31 Jul, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#8 IBM Mainframe Demise
IBM 1993: The Biggest Layoffs in US History
https://www.youtube.com/watch?v=GX4wJvQR4ok
1972, Learson tried (& failed) to block the bureaucrats, careerists,
and MBAs from destroying Watson culture/legacy, pg160-163, 30yrs of
management briefings 1958-1988
https://bitsavers.org/pdf/ibm/generalInfo/IBM_Thirty_Years_of_Mangement_Briefings_1958-1988.pdf
Future System disaster,
http://www.jfsowa.com/computer/memo125.htm
https://en.wikipedia.org/wiki/IBM_Future_Systems_project
https://people.computing.clemson.edu/~mark/fs.html
... from "Computer Wars: The Post-IBM World"
https://www.amazon.com/Computer-Wars-The-Post-IBM-World/dp/1587981394/
... and perhaps most damaging, the old culture under Watson Snr and Jr
of free and vigorous debate was replaced with *SYNCOPHANCY* and *MAKE
NO WAVES* under Opel and Akers. It's claimed that thereafter, IBM
lived in the shadow of defeat ... But because of the heavy investment
of face by the top management, F/S took years to kill, although its
wrong headedness was obvious from the very outset. "For the first
time, during F/S, outspoken criticism became politically dangerous,"
recalls a former top executive
... snip ...
I was blamed for online computer conferencing late 70s and early
80s. It really took off spring of 1981 when I distributed trip report
of visit to Jim Gray at Tandem. Only about 300 directly participated,
but claims that upwards of 25,000 were reading. Six copies of 300
pages were packaged in Tandem 3-ring binders and sent to the corporate
executive committee (folklore 5of6 wanted to fire me). From IBM Jargon
(Datamation reference was in original version):
https://havantcivicsociety.uk/wp-content/uploads/2019/05/ibmjarg.pdf
Tandem Memos - n. Something constructive but hard to control; a fresh
of breath air (sic). That's another Tandem Memos. A phrase to worry
middle management. It refers to the computer-based conference (widely
distributed in 1981) in which many technical personnel expressed
dissatisfaction with the tools available to them at that time, and
also constructively criticized the way products were [are]
developed. The memos are required reading for anyone with a serious
interest in quality products. If you have not seen the memos, try
reading the November 1981 Datamation summary.
... snip ...
There was executive summary and a summary of the executive summary;
from summary of summary:
• The perception of many technical people in IBM is that the company is
rapidly heading for disaster. Furthermore, people fear that this
movement will not be appreciated until it begins more directly to
affect revenue, at which point recovery may be impossible
• Many technical people are extremely frustrated with their management
and with the way things are going in IBM. To an increasing extent,
people are reacting to this by leaving IBM. Most of the contributors
to the present discussion would prefer to stay with IBM and see the
problems rectified. However, there is increasing skepticism that
correction is possible or likely, given the apparent lack of
commitment by management to take action
• There is a widespread perception that IBM management has failed to
understand how to manage technical people and high-technology
development in an extremely competitive environment
... snip ...
There were task forces to look at online computer conferencing
resulting in official IBM forum software and approved, moderated forum
discussions. Also a researcher was paid to study how I communicated;
sat in the back of my office for nine months taking notes on how I
communicated, also got copies of all my incoming and outgoing email
and logs of all instant messages. The result was used for IBM research
reports, conference talks and papers, books and Stanford Phd (joint
with language and computer AI, Winograd was advisor on computer AI
side).
I was told that with 5of6 of the executive committee wanting to fire
me, they would never be able to make me an IBM Fellow, but if I would
keep a low profile, they would divert funding my way as if I was. An
IBM Fellow became my 1st line, reporting to YKT (but left in San Jose,
having to commute to YKT a couple times a month).
I get funding for HSDT, T1 and faster computer links (terrestrial and
satellite) and battles with the communication group (60s had 2701 that
supported T1 links, 70s issues with VTAM capped controllers at
56kbits). Funding was predicated on being able to show some IBM
content, was able to find FSD S/1 T1 Zirpel cards (for gov. agency
2701s that were failing). Also working with NSF director and was
suppose to get $20M to interconnect the NSF supercomputer
centers. Then congress cuts the budget, some other things happened and
eventually there was RFP released (in part based on what we already
had running). NSF 28Mar1986 Preliminary Announcement
https://web.archive.org/web/20250428194307/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers
Program to provide Supercomputer cycles; the New Technologies Program
to foster new supercomputer software and hardware developments; and
the Networking Program to build a National Supercomputer Access
Network - NSFnet.
... snip ...
IBM internal politics was not allowing us to bid. The NSF director
tried to help by writing the company a letter (3Apr1986, NSF Director
to IBM Chief Scientist and IBM Senior VP and director of Research,
copying IBM CEO) with support from other gov. agencies ... but that
just made the internal politics worse (as did claims that what we
already had operational was at least 5yrs ahead of the winning bid),
as regional networks connect in, NSFnet becomes the NSFNET backbone,
precursor to modern internet.
Future System posts
https://www.garlic.com/~lynn/submain.html#futuresys
computer conferencing posts
https://www.garlic.com/~lynn/subnetwork.html#cmc
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
Internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
IBM downturn/downfall/breakup posts
https://www.garlic.com/~lynn/submisc.html#ibmdownfall
pension posts
https://www.garlic.com/~lynn/submisc.html#pension
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Mainframe HASP
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Mainframe HASP
Date: 03 Aug, 2026
Blog: Facebook
HASP was spooling program morphs into JES2 for MVS. ASP was
multi-system operation with shared DASD ... which morphs into JES3
(for MVS). My wife had been in the gburg JES group (headed by
Crabtree, one of the Houston HASP people). She was one of the ASP/JES3
"catchers" ... then co-author of "JESUS", features of JES2&JES3 that
the respective customers couldn't live w/o ... for various reasons
never came to fruition.
https://en.wikipedia.org/wiki/Houston_Automatic_Spooling_Priority
My wife was then talked into going to POK in charge of loosely-coupled
(mainframe for cluster) architecture
The internal network was primarily RSCS/VNET (initially morphing from
scientific center CP67-based wired-area network). The major MVS
support was JES2, code having come from HASP (at one time cards had
"TUCC" in cols 68-71). JES2 implementation still had the original
network node definitions that used spare entries in the 255-entry
psuedo device table (typically 160 to maybe 180) ... however the
internal network had rapidly passed 255 entries (and the JES2 code
would discard traffic where origin and/or destinition wasn't in local
table; by the time JES2 got around to increasing node definitions to
supporting 999, the internal network had already passed 1000.
HASP/JES2 networking evolved another problem. The network fields were
intermixed with job control fields. Minor release changes could result
in crashing the destination host MVS system. Eventually there were
RSCS/VNET emulated JES2 drivers that would recognize the immediately
connected destination JES2 release and reorganize the fields if
necssary (preventing destination MVS from crashing). There was
infamous case where the Hursley MVS systems were crashing and the
Hursley RSCS/VNET people were blamed because they hadn't updated the
RSCS/VNET emulated JES2 drivers to account for changes in San Jose MVS
JES2 systems.
In college took two credit hr intro to fortran/computers. At the end
of the semester, I was hired to re-implement 1401 MPIO (709
tape->tape, 1401 reader/punch/printer front-end) in assembler for
360/30. Univ. was getting 360/67 for tss/360 replacing 709/1401 and
got 360/30 temporarily until availability of 360/67s. Univ. shutdown
datacenter on weekends and I got the whole place dedicated (although
48hrs w/o sleep made Monday classes hard). I was given pile of
hardware & software manuals and got to design and implement my own
monitor, device drivers, interrupt handlers, error recovery, storage
management, etc ... and within a few weeks had 2000 card 360/30
assembler program
360/67 arrived within year of taking intro class and I was hired
fulltime responsible for os/360 (tss/360 never came to fruition). 709
did student fortran in less than second, but 360/67 os/360 took over
minute. I install HASP for MFT9.5 cutting time in half. I then start
redoing MFT11 SYSGEN STAGE2 carefully placing datasets and PDS members
optimizing arm seek and multi-track search cutting another 2/3rds to
12.9secs. 360/67 never got better than 709 until I install UofWaterloo
WATFOR, clocked on 360/67 at 20,000 cards/min (333 cards/sec
... student fortran tended to run 30-60cards/job).
Before I graduate, I was hired fulltime into small group in the Boeing
CFO office to help form the Boeing Computer Service (consolidate all
dataprocessing into independent business unit). I think Renton
datacenter largest in the world. Both Boeing and local IBM tell story
on day OS/360 was announced, Boeing gives the IBM marketing rep a 360
order, making him the highest paid IBM employee that year (days of
straight commission, following year, IBM introduces quotas). When I
graduate, I join IBM Cambridge Scientific Center (instead of staying
with CFO).
HASP/JES2, ASP/JES3, NJI/NJE posts
https://www.garlic.com/~lynn/submain.html#hasp
peer-coupled shared data (loosely-coupled) architecture posts
https://www.garlic.com/~lynn/submain.html#shareddata
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Terminal Support
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Terminal Support
Date: 04 Aug, 2026
Blog: Facebook
one of my hobbies after joining IBM was enhanced production operating
systems for internal datacenters and online, internal sales&marketing
support HONE systems was one of first (from 2741 days) and long time
customers (note: when FACEBOOK 1st moved into Silicon Valley, it was
new bldg built next door to the former consolidated US HONE
datacenter). Channel attached 3272/3277 had .086sec hardware
response. 3278 period started to see studies showing .25sec response
improved productivity. For 3278, lots of electronics moved back into
3274 (reducing 3278 manufacturing cost) driving up coax protocol
latency to .3sec-.5sec. (depending on amount of data) I had lots of
systems that were getting .11sec interactive system response
(resulting .086+.11=.196sec total elapsed response, within .25sec
target, which 3278s were NEVER able to meet). Letters to 3278 product
administrator complaining got responses that 3278 wasn't for
interactive computing, but data entry. MVS users never really noticed
the 3277/3278 difference since it was rare for MVS to even see 1sec
system response. Then IBM/PC 3277 emulation cards had 4-5 times the
upload/download throughput of 3278 emulation cards.
... also 1980, STL (since renamed SVL) was bursting at the seams, and
300 people (and 3270s) from IMS group were being moved to offsite bldg
with dataprocessing service back to STL. They had tried "remote 3270"
but found the human factors unacceptable. I get con'ed into doing
channel-extender support that placed 3270 channel-attached controllers
at the offsite bldg; resulting in no perceptible human factors
difference with inside STL. Unattended side-effects was that the
systems with channel-extender 3270 controllers saw a 10-15%
improvement in system throughput. STL had been distributing
channel-attached 3270 controllers across the channels with 3330
DASD. The channel-extenders significantly reduced the 3270 controller
(168) channel busy (for same amount of 3270 I/O) ... and they were
considering placing all 3270 controllers on channel-extenders.
trivia:
Took two credit hr intro to fortran/computers. At the end of the
semester, I was hired to re-implement 1401 MPIO (709 tape->tape, 1401
reader/punch/printer front-end) in assembler for 360/30. Univ. was
getting 360/67 for tss/360 replacing 709/1401 and got 360/30
temporarily until availability of 360/67s. Univ. shutdown datacenter
on weekends and I got the whole place dedicated (although 48hrs w/o
sleep made Monday classes hard). I was given pile of hardware &
software manuals (learn assembler and hardware) and got to design and
implement my own monitor, device drivers, interrupt handlers, error
recovery, storage management, etc ... and within a few weeks had 2000
card 360/30 assembler program. Assembler option for two modes; 1)
stand-alone, IPLed with BPS loader and 2) OS/360 macros, GET/PUT,
OPEN/CLOSE, etc. Sometimes when I came in on weekend, production had
finished early and everything was dark/shutdown ... and 360/30 powerup
wouldn't work. Some manuals and experimenting, I found if I put all
the controllers in CE-mode, I could power on the 360/30, then power on
individual controllers and take them out of CE-mode.
The 360/67 arrived with-in a year of taking intro class and I was
hired fulltime responsible for OS/360 (TSS/360 not coming to
production). 709 did student fortran in less than second. 360/67 (as
360/65) os/360 (MFT9.5) took over a minute, I install HASP cutting
time in half. MFT11, I start redoing SYSGEN STAGE2 to carefully place
datasets and PDS members to optimize arm seek and (PDS directory)
multi-track search, cutting another 2/3rds to 12.9secs. OS/360 never
got better than 709 until I install Univ Waterloo WATFOR; on 360/65
WATFOR clocked at 20,000 statements/min (333/sec), student fortran
tended to run 30-60 statements.
Then IBM CSC came out to install (virtual machine) CP/67 (3rd after
CSC itself and MIT Lincoln Labs, precursor to VM370) and I mostly get
to play with it during my weekend 48hr window. I then spend a few
months rewriting pathlengths for running OS/360 in virtual
machine. Bare machine test ran 322secs ... initially 856secs (CP67 CPU
534secs). After a few months I had CP67 CPU down from 534secs to
113secs. I then start rewriting the dispatcher/scheduler , (dynamic
adaptive resource manager/default fair share scheduling policy),
paging, adding ordered seek queuing (from FIFO) and mutli-page
transfer channel programs (from FIFO and optimized for
transfers/revolution, getting 2301 paging drum from 70-80 4k
transfers/sec to channel transfer peak of 270). Six months after univ
initial CP/67 install, CSC was giving one week class in LA. I arrive
on Sunday afternoon and asked to teach the class, it turns out that
the IBMers that were going to teach it had resigned the Friday before
to join one of the 60s CSC CP67 commercial online spin-offs.
CP/67 arrived with 1052&2741 terminal support and auto-terminal ident,
capable switching terminal type scanner type for each port. Univ. also
had TTY33&35 (paper tape) and I add TTY ASCII support integrated with
auto-terminal type (ASCII port scanner had arrived in a Heathkit box;
TTY required much more finger strength than IBM selectric
typeball-based 1052&2741). I then want to have single dial-in number
(hunt group) for all terminals. Didn't quite work, IBM had hard-wired
port line speed ... so we start a clone terminal controller. Build a
IBM channel interface board for Interdata/3 programmed to emulate IBM
controller with the addition for auto-baud. Then upgraded with
Interdata/4 for channel interface and cluster of Interdata/3s for port
interfaces. Interdata (and later Perkin-Elmer) sell them as clone
controllers. https://en.wikipedia.org/wiki/Interdata
https://en.wikipedia.org/wiki/Perkin-Elmer#Computer_Systems_Division
... and four of us are written up responsible for (some part of) IBM
clone controller business
trivia2:
Mid-80s, communication group was fighting off releasing mainframe
TCP/IP (for VM370), when they lost they said that since they had
corporate ownership of everything that crossed datacenter walls, it
had to be released through them. What shipped got aggregate
44kbytes/sec using nearly whole 3090 processor. I then added RFC1044
support and in some tuning tests at Cray Research between Cray and
4341, got sustained 4341 channel throughput using only modest 4341 CPU
(something like 500 times improvement in bytes moved per instruction
executed). The VM370 TCP/IP support was released for MVS (by
simulating some VM370 "diagnose" instruction).
Late 80s, univ study found BSD (Reno/Tahoe) UNIX TCP pathlengths had
been clocked at 5k instructions while VTAM LU6.2 pathlengths was 160k
instructions.
Later in the early 90s, communication group hired silicon valley
contractor to add TCP/IP support to VTAM. What he 1st demo'ed had TCP
running much faster than LU6.2. He was then told that everybody knows
that a "proper" TCP/IP implementation is much slower than LU6.2 and
they would only be paying for a "proper" implementation.
trivia3:
1988, the IBM branch office asks if I can help LLNL (national lab)
with standardization of some serial stuff they were working with
... which quickly becomes fibre-channel standard ("FCS", including
some stuff I had done in 1980, initially 1gbit transfer, full-duplex,
aggregate 200mbyte/sec, and non-blocking 64-port switch). Then IBM POK
mainframe releases some serial (when it was already obsolete) as
ESCON, initially 10mbyte/sec. Then some POK engineers become involved
with "FCS" and define a heavy-weight protocol that drastically cuts
native throughput, eventually ships as FICON.
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
IBM CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
online commercial CP67 service
https://www.garlic.com/~lynn/submain.html#online
clone controller posts
https://www.garlic.com/~lynn/submain.html#360pcm
channel-extender posts
https://www.garlic.com/~lynn/submisc.html#channel.extender
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
XTP/HSP posts
https://www.garlic.com/~lynn/subnetwork.html#xtphsp
FCS and/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM 370 CISC Chip
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM 370 CISC Chip
Date: 05 Aug, 2026
Blog: Facebook
2nd half 70s, an effort was started to use 801/RISC for many of IBM
microprocessors, AS/400, 4361, 4381, controllers, etc. I got asked to
help with document that showed it was possible to do 370 CISC chip
that implemented most of 370 directly in circuits. 370/138, 370/148,
4331, 4341 were done in microprocessors that avg. 10 microcode
instructions for every 370 instruction. Earlier for 138/148 I had been
asked to help with ECPS, moving highest executed operating system
paths directly into native microcode rather than 370 instruction on
approx. 1:1 basis.
trivia: initial analysis for ECPS ... 6kbytes of operating system 370
instruction paths accounted for 79.55% of operating system execution
... ECPS getting 10 times speedup:
https://www.garlic.com/~lynn/94.html#21
trivia2: i86 makers did something similar to ECPS in mid-90s ... where
hardware layer dynamically ("on the fly") translated i86 instructions
into risc micro-ops for actual execution ... supporting out-of-order
execution and multiple executions.
1999 industry benchmarks (number program iterations compared to
industry MIPS/BIPS reference platform, not physical count of their
instructions)
IBM PowerPC 440: 1,000MIPS
Pentium3: 2,054MIPS (twice PowerPC 440)
2010 (same) benchmarks (compared to industry MIPS/BIPS) reference
platform)
z196, 80 cores, 50BIPS, 625MIPS/core
E5-2600, 2 Intel XEON, 8-core chips, 16 cores, 500BIPS, 32BIPS/core
801/risc, iliad, romp, rios, pc/rt, rs/6000, power, power/pc posts
https://www.garlic.com/~lynn/subtopic.html#801
--
virtualization experience starting Jan1968, online at home since Mar1970
Remember when discussions were worth reading?
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: Remember when discussions were worth reading?
Newsgroups: alt.folklore.computers
Date: Fri, 07 Aug 2026 16:00:55 -1000
Some of the MIT CTSS/7094
https://en.wikipedia.org/wiki/Compatible_Time-Sharing_System
people went to the 5th flr and MULTICS
https://en.wikipedia.org/wiki/Multics
others went to the 4th flr and IBM Cambridge Science Center
https://en.wikipedia.org/wiki/Cambridge_Scientific_Center
CTSS RUNOFF
https://en.wikipedia.org/wiki/TYPSET_and_RUNOFF
was redone for CMS as "SCRIPT"
One of the 1969 GML (GML tag processing added to SCRIPT, later morphs
into SGML and HTML) inventors, CSC had originally hired to promote the
Scientific Center wide-area network;
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for the
project, something I was allowed to do part-time because of my knowledge
of the user requirements. My real job was to encourage the staffs of the
various scientific centers to make use of the CP-67-based Wide Area
Network that was centered in Cambridge.
... snip ...
which becomes RSCs/VNET, used for the internal corporate network (larger
than arpanet/internet from just about the beginning until sometimed
mid/late 80s about the time forced to convert SNA/VTAM) ... also used
for the corporate sponsored univ. BITNET
https://en.wikipedia.org/wiki/BITNET
... CSC CP/67 was precursor to VM370. Edson was responsible for CP67
wide-area network
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which paved
the way to the Internet as we know it today.
... snip ...
SJMerc article about Edson (he passed aug2020) and "IBM'S MISSED
OPPORTUNITY WITH THE INTERNET" (gone behind paywall but lives free at
wayback machine)
https://web.archive.org/web/20000124004147/http://www1.sjmercury.com/svtech/columns/gillmor/docs/dg092499.htm
Also from wayback machine, some additional (IBM missed internet)
references from Ed's website
https://web.archive.org/web/20000115185349/http://www.edh.net/bungle.htm
2nd half 70s, Ed and I transfer out to SJR on the west coast and I get
to wander around IBM mainframe datacenters in silicon valley
... including TYMSHARE
TYMSHARE
https://en.wikipedia.org/wiki/Tymshare
in Aug1976, started offering their (IBM mainframe) VM370/CMS-based
computer conferencing, free to the (user group) SHARE organization
https://en.wikipedia.org/wiki/SHARE_(computing)
archives here:
http://vm.marist.edu/~vmshare
I cut a deal with TYMSHARE for a monthly tape dump of all VMSHARE files
for puting up on the internal network and systems. Late 70s and early
80s, I'm also blamed for online computer conferencing on the internal
network. It really took off spring 1981 when I distributed trip report
to visit Jim Gray at Tandem. We packaged up six copies of 300 pages in
Tandem 3-ring binders and set them to corperate executive committee
(folklore is that 5of6 wanted to fire me).
Early 80s, I got "HSDT", T1 and faster computer links (terrestrial and
satellite) and battles with the communication group (60s had 2701 that
supported T1 links, 70s issues with VTAM capped controllers at
56kbits). Funding was predicated on being able to show some IBM content,
was able to find FSD S/1 T1 Zirpel cards (for gov. agency 2701s that
were failing). Also working with NSF director and was suppose to get
$20M to interconnect the NSF supercomputer centers. Then congress cuts
the budget, some other things happened and eventually there was RFP
released (in part based on what we already had running). NSF 28Mar1986
Preliminary Announcement
https://web.archive.org/web/20250428194307/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers Program
to provide Supercomputer cycles; the New Technologies Program to foster
new supercomputer software and hardware developments; and the Networking
Program to build a National Supercomputer Access Network - NSFnet.
...
IBM internal politics was not allowing us to bid. The NSF director tried
to help by writing the company a letter (3Apr1986, NSF Director to IBM
Chief Scientist and IBM Senior VP and director of Research, copying IBM
CEO) with support from other gov. agencies ... but that just made the
internal politics worse (as did claims that what we already had
operational was at least 5yrs ahead of the winning bid), as regional
networks connect in, NSFnet becomes the NSFNET backbone, precursor to
modern internet.
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
internal network
https://www.garlic.com/~lynn/subnetwork.html#internalnet
online computer conferencing posts
https://www.garlic.com/~lynn/subnetwork.html#cmc
commercial online (virtual machine) service posts
https://www.garlic.com/~lynn/submain.html#online
hsdt posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Publication
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Publication
Date: 07 Aug, 2026
Blog: Facebook
When I graduate, I join IBM Cambridge Scientific Center (instead of
staying with Boeing CFO). Late 70s, I transfer out to SJR on the west
coast and get to wander around IBM (and non-IBM) datacenters in
silicon valley, including disk bldg14 (engineering) and bldg15
(product test) across the street. They were running, 7x24,
pre-scheduled, stand-alone testing and mentioned they had recently
tried MVS (but it had 15min MTBF in that environment, requiring
re-ipl). I offer to rewrite I/O supervisor making it bullet-proof and
never fail, allowing any amount of on-demand concurrent testing,
greatly improving productivity.
Not for external release, I write an (internal) research report on the
I/O Reliability Enhancement work and happen to mention MVS 15min MTBF, bringing down
the wrath of the MVS organization on my head (was in told in part,
that there was no way I would get any kind of award for the work,
given the MVS organization opposition).
I write a number of additional publications that were being handled by
the senior San Jose GPD publication editor ... he said that research
management was constantly requiring additional document editing.
In late 70s and early 80s, I had been blamed for online computer
conferencing on the internal network. It really took off the spring of
1981 when I distribute trip report of visit to Jim Gray at
Tandem. Only about 300 directly participated but claims 25,000 were
reading. We printed six copies of 300 pages with executive summary and
summary of the summary, packaged in Tandem 3-ring binders and sent to
the corporate executive committee (folklore is that 5of6 wanted to
fire me). Nearly decade later, the senior GPD publication editor was
retiring and cleaning out his files and sent working copies of all my
draft research reports (that he could never get approved for
publication).
posts getting to play disk engineer in bldgs 14&15
https://www.garlic.com/~lynn/subtopic.html#disk
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
online computer conferencing posts
https://www.garlic.com/~lynn/subnetwork.html#cmc
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Unbundling and HONE
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Unbundling and HONE
Date: 08 Aug, 2026
Blog: Facebook
IBM 23jun1969 unbundling announce started to charge for (application)
software (managed to make case kernel software was still free), SE
services, maint, etc. SE training used to include part of SE group
onsite at customers, but IBM couldn't figure out how not to charge for
trainee SEs onsite. HONE was the result, multiple CP67 datacenters
around the US, that branch office people could dial-in (2741s) and
practice with guest operating systems running in virtual
machines (after joining IBM, one of my hobbies was enhanced
production operating systems for internal datacenters and HONE was one
of the first and long time customer).
CSC also ports APL\360 to CMS as CMS\APL (included API that allowed
invoking system services like file i/o) and HONE started offering
(CP67) CMS\APL-based online sales&marketing support applications
... which came to dominate all HONE activity (guest operating system
practice just withered away). HONE migrated to VM370/CMS and all US
HONE datacenters were consolidated in Palo Alto (trivia: when FACEBOOK
1st moves into silicon valley, it was into new bldg built next door to
the former US HONE datacenter), while other HONE datacenters were
starting to sprout up all over the world (HONE becoming the largest
APL use in the world).
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
23jun1969 unbundling announce posts
https://www.garlic.com/~lynn/submain.html#unbundle
HONE (& APL) posts
https://www.garlic.com/~lynn/subtopic.html#hone
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM 370/125
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM 370/125
Date: 09 Aug, 2026
Blog: Facebook
I got con'ed into designing and implementing 5-processor 370/125 (same
time Endicott had me doing ECPS for 138&148). 115/125 had
9-position memory bus. 115 had all microprocessors the same (emulated
controllers and microprocessor with 370 microcode clocked at
80KIPS). 125 was the same, but the 370 microprocessor was 50% faster
(clocked at 120KIPS) than the other microprocessors. The 125
multiprocessor version would have two to five microprocessors with 370
microcode.
Then Endicott complained that the 5-CPU 125 would overlap the 148 and
escalated to corporate (I had to attend and argue both sides) and the
multiprocessor 125 got canceled. I was then talked into helping with
the 16-CPU 370 (and we con the 3033 processor engineers into working
on it in their spare time, a lot more interesting than remapping 168
logic to 20% faster chips).
A few yrs later, one the IBMers I worked with had left and was working
for Apple on the Cray machine. It had a large, high-performance
graphics display on Cray channel and programmed to emulate Apple
display. He could simulate screen resolution, screen update time,
system response, etc.
5-CPU 370/125 posts
https://www.garlic.com/~lynn/submain.html#bounce
SMP, tightly-coupled, shared-memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
some recent posts mentioning both 16-CPU 370 and 138/148 ECPS
https://www.garlic.com/~lynn/2026b.html#54 VM4341 Cluster Supercomputing and Departmental Distributed
https://www.garlic.com/~lynn/2026b.html#4 IBM Virtual Machine
https://www.garlic.com/~lynn/2026.html#92 CP/67 and VM/370
https://www.garlic.com/~lynn/2026.html#71 IBM 3033
https://www.garlic.com/~lynn/2026.html#25 Amdahl Computers
https://www.garlic.com/~lynn/2026.html#10 4341, cluster supercomputing, distributed computing
https://www.garlic.com/~lynn/2025e.html#62 IBM Mainframe Projects
https://www.garlic.com/~lynn/2025d.html#61 Amdahl Leaves IBM
https://www.garlic.com/~lynn/2025c.html#49 IBM And Amdahl Mainframe
https://www.garlic.com/~lynn/2025b.html#118 IBM 168 And Other History
https://www.garlic.com/~lynn/2025b.html#58 IBM Downturn, Downfall, Breakup
https://www.garlic.com/~lynn/2025b.html#46 POK High-End and Endicott Mid-range
https://www.garlic.com/~lynn/2025.html#43 Multics vs Unix
https://www.garlic.com/~lynn/2024g.html#112 Dataprocessing Innovation
https://www.garlic.com/~lynn/2024e.html#129 IBM 4300
--
virtualization experience starting Jan1968, online at home since Mar1970
NSF Supercomputer and Network
From: Lynn Wheeler <lynn@garlic.com>
Subject: NSF Supercomputer and Network
Date: 12 Aug, 2026
Blog: Facebook
I had HSDT effort (t1 and faster computer links, both terrestrial and
satellite) starting in the early 80s and was working with NSF director
and suppose to get $20M to interconnect the NSF supercomputer
centers. UCB/Berkeley was suppose to get grant for NSF supercomputer
center and I was asked to give presentations at UCB about the
interconnect. Turns out UC Regents building plan had San Diego getting
the next new building ... and it became the San Diego Supercomputer
center (since the NSF grant went to UC system, NSF didn't actually
have any direct control over which campus it was used for), operated
by General Atomics.
Then congress cuts the budget, some other things happened and
eventually there was RFP released (in part based on what we already
had running). NSF 28Mar1986 Preliminary Announcement
https://web.archive.org/web/20250428194307/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers
Program to provide Supercomputer cycles; the New Technologies Program
to foster new supercomputer software and hardware developments; and
the Networking Program to build a National Supercomputer Access
Network - NSFnet.
...
IBM internal politics was not allowing us to bid. The NSF director
tried to help by writing the company a letter (3Apr1986, NSF Director
to IBM Chief Scientist and IBM Senior VP and director of Research,
copying IBM CEO) with support from other gov. agencies ... but that
just made the internal politics worse (as did claims that what we
already had operational was at least 5yrs ahead of the winning bid),
as regional networks connect in, NSFnet becomes the NSFNET backbone,
precursor to modern internet.
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
some posts mentioning HSDT, NSFNET, NSF supercomputer, regents, UCB,
UCSD
https://www.garlic.com/~lynn/2024g.html#73 Early Email
https://www.garlic.com/~lynn/2024f.html#73 IBM 2250 Hypertext Editing System
https://www.garlic.com/~lynn/2024b.html#51 IBM Token-Ring
https://www.garlic.com/~lynn/2021g.html#61 IBM HSDT & HA/CMP
https://www.garlic.com/~lynn/2019e.html#88 5 milestones that created the internet, 50 years after the first network message
--
virtualization experience starting Jan1968, online at home since Mar1970
NSF Supercomputer and Network
From: Lynn Wheeler <lynn@garlic.com>
Subject: NSF Supercomputer and Network
Date: 13 Aug, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#17 NSF Supercomputer and Network
There was internal IBM projects that had done mainframe TCP/IP in
VS/Pascal (IBM Los Gatos lab was using Metaware's TWS for various VLSI
projects and had used it for mainframe Pascal). The communication
group was fighting the mainframe TCP/IP release. When they lost, they
said that since they had corporate responsibility for everything that
crossed datacenter walls, it had to be released through them. What
shipped got aggregate 44kbytes/sec using nearly whole 3090
processor. I then added RFC1044 support and in some tuning tests at
Cray Research between Cray and IBM 4341 (slightly more than one MIPs,
competed with DEC VAX in mid-range market), got sustained 4341 channel
throughput using modest amount of 4341 CPU (something like 500 times
more bytes moved per instruction executed).
RFC1044 posts
https://www.garlic.com/~lynn/subnetwork.html#1044
Internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
1988, IBM Branch Office asked if I could help LLNL with standardizing
some serial stuff they were playing, which quickly becomes
fibre-channel standard (FCS, including some stuff I had done in 1980,
initially 1gbit transfer, full-duplex, aggregate 200mbytes/sec). Same
time, was also asked to help SLAC/Gustavson with SCI standard (Data
General and Sequent did 64 4-i486 boards, 256 processor SCI
NUMA). After leaving IBM did some consulting for Steve Chen (had done
Cray YMP), who was then Sequent CTO (before IBM bought it and shut it
down).
FCS and/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
Also 1988, got HA/6000 project, initially for NYTimes to move their
newspaper system (ATEX) off DEC VAXCluster to RS/6000, I rename it
HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing technical/scientific cluster scale-up with national
labs (LANL, LLNL, NCAR, etc) and commercial cluster scale-up with
RDBMS vendors (Oracle, Sybase, Ingres, Informix) with VAXCluster
support in same source base with Unix. Ported LLNL's UNICOS LINCS and
NCAR's filesystem (spun off as Mesa Archival) to HA/CMP.
IBM S/88 (relogo'ed Stratus) Product Administrator started taking us
around to their customers and also had me write a section for the
corporate continuous availability strategy document (it gets pulled
when both AS400/Rochester and mainframe/POK complain they couldn't
meet requirements). Had coined disaster survivability and
geographic survivability (as counter to disaster/recovery) when out
marketing HA/CMP. One of the visits to 1-800 bellcore development
showed that S/88 would use a century of downtime in one software
upgrade, while HA/CMP had a couple extra "nines" (compared to S/88).
One of the first HA/CMP customer installs was new Indian Reservation
Casino in Connecticut, was suppose to have week of testing before
opening ... but after 24hrs, they decided to open the doors (based on
projected revenue; at the time was largest in the US, still one of the
largest in the country)
https://en.wikipedia.org/wiki/Foxwoods_Resort_Casino#Debt_default
Early Jan92, there was HA/CMP meeting with Oracle CEO and IBM/AWD
executive Hester tells Ellison that we would have 16-system clusters
by mid92 and 128-system clusters by ye92. Mid-jan92, I update FSD on
HA/CMP work with national labs and FSD decides to go with HA/CMP for
federal supercomputers. By end of Jan92, we are told that cluster
scale-up is being transferred to Kingston for announce as IBM
Supercomputer (SP, technical/scientific *ONLY*) and we aren't allowed
to work with anything that has more than four systems (we leave IBM a
few months later). A couple weeks later, 17feb1992, Computerworld news
... IBM establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
some posts mentioning mainframe tcp/ip, pascal, metware tws, los gatos
https://www.garlic.com/~lynn/2025.html#114 IBM 370 Virtual Memory
https://www.garlic.com/~lynn/2024c.html#114 Disconnect Between Coursework And Real-World Computers
https://www.garlic.com/~lynn/2024.html#70 IBM AIX
https://www.garlic.com/~lynn/2024.html#8 Niklaus Wirth 15feb1934 - 1jan2024
https://www.garlic.com/~lynn/2023g.html#68 Assembler & non-Assembler For System Programming
https://www.garlic.com/~lynn/2023c.html#98 Fortran
https://www.garlic.com/~lynn/2022h.html#40 Mainframe Development Language
https://www.garlic.com/~lynn/2022g.html#6 "In Defense of ALGOL"
https://www.garlic.com/~lynn/2022f.html#13 COBOL and tricks
https://www.garlic.com/~lynn/2021c.html#95 What's Fortran?!?!
https://www.garlic.com/~lynn/2017e.html#24 [CM] What was your first home computer?
https://www.garlic.com/~lynn/2015g.html#52 [Poll] Computing favorities
https://www.garlic.com/~lynn/2013m.html#36 Quote on Slashdot.org
https://www.garlic.com/~lynn/2013l.html#59 Teletypewriter Model 33
https://www.garlic.com/~lynn/2010n.html#54 PL/I vs. Pascal
https://www.garlic.com/~lynn/2009l.html#36 Old-school programming techniques you probably don't miss
https://www.garlic.com/~lynn/2008j.html#77 CLIs and GUIs
some posts mentioning SCI, Sequent, Steve Chen, NUMA
https://www.garlic.com/~lynn/2026c.html#7 SCI, FCS, HA/CMP
https://www.garlic.com/~lynn/2024e.html#90 Mainframe Processor and I/O
https://www.garlic.com/~lynn/2024.html#54 RS/6000 Mainframe
https://www.garlic.com/~lynn/2022g.html#91 Stanford SLAC (and BAYBUNCH)
https://www.garlic.com/~lynn/2021i.html#16 A brief overview of IBM's new 7 nm Telum mainframe CPU
https://www.garlic.com/~lynn/2019.html#32 Cluster Systems
https://www.garlic.com/~lynn/2018d.html#57 tablets and desktops was Has Microsoft
https://www.garlic.com/~lynn/2017c.html#49 The ICL 2900
https://www.garlic.com/~lynn/2014.html#71 the suckage of MS-DOS, was Re: 'Free Unix!
https://www.garlic.com/~lynn/2013n.html#50 'Free Unix!': The world-changing proclamation made30yearsagotoday
https://www.garlic.com/~lynn/2013h.html#6 The cloud is killing traditional hardware and software
https://www.garlic.com/~lynn/2012p.html#13 AMC proposes 1980s computer TV series Halt & Catch Fire
https://www.garlic.com/~lynn/2010i.html#61 IBM to announce new MF's this year
--
virtualization experience starting Jan1968, online at home since Mar1970
NSF Supercomputer and Network
From: Lynn Wheeler <lynn@garlic.com>
Subject: NSF Supercomputer and Network
Date: 14 Aug, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#17 NSF Supercomputer and Network
https://www.garlic.com/~lynn/2026c.html#18 NSF Supercomputer and Network
2nd half 70s, I transfer out to IBM SJR on the west coast and get to
wander around datacenters in silicon valley, including disk
bldg14/engineering and bldg15/product test. They were doing 7x24,
prescheduled, stand-alone testing and mentioned that they had recently
tried MVS, but it had 15min MTBF (in that environment, require manual
re-ipl). I offer to rewrite I/O supervisor making it bullet proof and
never fail, allowing any amount of on-demand, concurrent testing,
greatly improving productivity.
Bldg15 then gets 1st engineering 3033 outside POK processor
engineering. Testing was only taking percent or two of CPU, so we
scrounge up 3830 controller and 3330 string for our own private online
service. Then summer 1978, bldg15 gets an engineering 4341 (a year
before 1st customer ship). Jan1979, branch office hears about it and
cons me into doing (CDC6600 Fortran) benchmark for national lab that
was looking at getting 70 VM4341s for compute farm (4341 benchmarked
about same as decade old 6600, but 70, sort of the leading edge of the
coming cluster supercomputer tsunami).
A decade of VAX sales, sliced and diced by model, year, us/non-us
https://web.archive.org/web/20260305143949/https://www.garlic.com/~lynn/2002f.html#0
4300s sold in similar numbers as VAX into mid-range market for small
unit orders. Big difference was large corporations with orders of
hundreds of VM4300s at a time, for placing out in departmental areas
(sort of the leading edge of the coming distributed computing
tsunami). Inside IBM conference rooms were becoming scarce, so many
being converted to departmental/distributed computing.
posts getting to play disk engineer
https://www.garlic.com/~lynn/subtopic.html#disk
--
virtualization experience starting Jan1968, online at home since Mar1970
ACP/TPF
From: Lynn Wheeler <lynn@garlic.com>
Subject: ACP/TPF
Date: 16 Aug, 2026
Blog: Facebook
3081 was going to be multiprocessor only, 2-CPU 3081D had smaller
aggregate MIPs than 1-CPU Amdahl. IBM then doubles 3081 processor
cache sizes for 2-CPU 3081K, bringing aggregate MIPS up to about the
same as Amdahl 1-CPU. However ACP/TPF didn't have shared-memory
multiprocessor support, only supported 1-CPU system ... and IBM was
concerned that the whole ACP/TPF market would move to Amdahl 1-CPU
systems. Then some very unnatural things done to VM/370 to improve
ACP/TPF throughput running in 1-CPU virtual machine .... however it
degraded the system throughput by 10-15% for nearly every other
customer VM370 multiprocessor system. Eventually IBM did 3083 (for
ACP/TPF) where one 3081 CPU was removed (but it never matched Amdahl
1-CPU systems). I got brought into some number of these other
customers (but IBM wouldn't allow me to regress the VM370
multiprocessor changes for ACP/TPF).
Later in 80s, my wife did short stint as chief architect for Amadeus
(built off the old Eastern 370/195 "System One"). However she sided
with Europe about X.25 and the IBM communication group got her
replaced. It didn't do them much good, Europe went with X.25 anyway
and the communication group replacement was replaced.
Mid-90s (after leaving IBM), I was brought into the largest airline
res system to look at the 10 impossible things they couldn't do. They
started with "ROUTES" (represented about 25% of the workload and took
about 1/4 of organization people). Walked me through the impossible
things and I got softcopy of full OAG (all commercial scheduled flts
in the world). I went away and after a couple weeks had a C-language
RS/6000 that implemented existing ROUTE function that ran about 100
times faster. I then added support for the impossible things ... which
slowed it down to about ten times faster ... and took it back and let
them test it. Benchmarks showed a cluster of ten RS6000/990s would be
able to handle every ROUTE transaction for every scheduled airline in
the world (and reduce "ROUTES" support head count from around 800 to
about 40).
trivia: less than decade later cellphone (XSCALE) processor had more
MIPS than aggregate of the ten 990s.
SMP, tightly-coupled, shared memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
a few posts mentioning doing work on ACP/TPF "ROUTES"
https://www.garlic.com/~lynn/2026b.html#59 Mainframe, Workstation, Cellphone CPU
https://www.garlic.com/~lynn/2026b.html#45 STL Channel Extender
https://www.garlic.com/~lynn/2025e.html#55 IBM ACP/TPF
https://www.garlic.com/~lynn/2024e.html#92 IBM TPF
https://www.garlic.com/~lynn/2023g.html#90 Has anybody worked on SABRE for American Airlines
https://www.garlic.com/~lynn/2023g.html#74 MVS/TSO and VM370/CMS Interactive Response
https://www.garlic.com/~lynn/2023.html#96 Mainframe Assembler
https://www.garlic.com/~lynn/2021i.html#77 IBM ACP/TPF
https://www.garlic.com/~lynn/2021i.html#76 IBM ITPS
https://www.garlic.com/~lynn/2016.html#58 Man Versus System
https://www.garlic.com/~lynn/2015f.html#5 Can you have a robust IT system that needs experts to run it?
--
virtualization experience starting Jan1968, online at home since Mar1970
CMSBACK, ADSM, TSM
From: Lynn Wheeler <lynn@garlic.com>
Subject: CMSBACK, ADSM, TSM
Date: 17 Aug, 2026
Blog: Facebook
I had done CMSBACK in 1979 ... 1st used in SJR (before move up the
hill to Almaden) and then in US Consolidated (IBM branch office
marketing&sales support) HONE datacenters up in Palo Alto (one of my
hobbies after joining IBM was enhanced production operating systems
for internal datacenters and HONE was one of the 1st and long time
customer) then some number other IBM locations. After my initial
internal release, I worked with SJR coworker ... who left IBM and did
VMBACKUP release 2 in 1982 for VM Software. Later in the 80s, some of
the Almaden people added PC and workstation clients and it was
released as WDSF (workstation datasave facility), which morphed into
ADSM and then TSM
https://en.wikipedia.org/wiki/IBM_Tivoli_Storage_Manager
https://en.wikipedia.org/wiki/IBM_Tivoli_Storage_Manager#History
TSM descended from a project done at IBM's Almaden Research Center
around 1988 to back up VM/CMS systems. The first product that emerged
was Workstation Data Save Facility (WDSF). WDSF's original purpose was
to back up PC/DOS, OS/2, and AIX workstation data onto a VM/CMS (and
later MVS) server. WDSF morphed into ADSTAR Distributed Storage
Manager (ADSM) and was re-branded Tivoli Storage Manager in 1999.
...
IBM's GPD was rebranded as ADSTAR as part of reorganizing IBM into the
13 "baby blues" in prepartion for breaking up the company (1992, IBM
had one of the largest losses in the history of US companies).
https://web.archive.org/web/20101120231857/http://www.time.com/time/magazine/article/0,9171,977353,00.html
https://content.time.com/time/subscriber/article/0,33009,977353-1,00.html
we had already left IBM but get a call from the bowels of Armonk
asking if we could help with the breakup of the company. Before we get
started, the board brings in the former Amex president for CEO to try
and save the company, who (somewhat) reverses the breakup and uses
some of the same techniques used at RJR (gone 404, but lives on at
wayback)
https://web.archive.org/web/20181019074906/http://www.ibmemployee.com/RetirementHeist.shtml
While former president of Amex reverses the original breakup, much of
the hardware portion of ADSTAR was sold off ... and much of ADSTAR
software was transferred to other IBM organizations.
Early 80s, I got HSDT project (and responsibility for CMSBACK was
transferred to other SJR people), T1 and faster computer links
(terrestrial and satellite) and battles with the communication group
(60s had 2701 that supported T1 links, 70s issues with VTAM capped
controllers at 56kbits). Also working with NSF director and was
suppose to get $20M to interconnect the NSF supercomputer centers. NSF
was giving a grant to UC for Berkeley Supercomputer Center and I was
asked to give UCB presentations on the interconnect. However the UC
Regents had UCSB getting the next new bldg and it becomes the San
Diego Supercomputer Center instead (NSF had no control).
Then congress cuts the budget, some other things happened and
eventually there was RFP released (in part based on what we already
had running). NSF 28Mar1986 Preliminary Announcement
https://web.archive.org/web/20250428194307/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers
Program to provide Supercomputer cycles; the New Technologies Program
to foster new supercomputer software and hardware developments; and
the Networking Program to build a National Supercomputer Access
Network - NSFnet.
...
IBM internal politics was not allowing us to bid. The NSF director
tried to help by writing the company a letter (3Apr1986, NSF Director
to IBM Chief Scientist and IBM Senior VP and director of Research,
copying IBM CEO) with support from other gov. agencies ... but that
just made the internal politics worse (as did claims that what we
already had operational was at least 5yrs ahead of the winning bid),
as regional networks connect in, NSFnet becomes the NSFNET backbone,
precursor to modern internet.
some recent posts mentioning CMSBACK:
https://www.garlic.com/~lynn/2025b.html#91 IBM AdStar
https://www.garlic.com/~lynn/2025.html#56 IBM CMSBACK, WDSF, ADSM, TSM
https://www.garlic.com/~lynn/2024e.html#87 IBM San Jose
https://www.garlic.com/~lynn/2024e.html#75 IBM San Jose
https://www.garlic.com/~lynn/2024b.html#7 IBM Tapes
https://www.garlic.com/~lynn/2023e.html#98 Mainframe Tapes
https://www.garlic.com/~lynn/2023e.html#81 Storage Management
https://www.garlic.com/~lynn/2023c.html#104 IBM Term "DASD"
https://www.garlic.com/~lynn/2023c.html#55 IBM VM/370
https://www.garlic.com/~lynn/2022c.html#85 VMworkshop.og 2022
https://www.garlic.com/~lynn/2022c.html#32 CMSBACK & VMFPLC
https://www.garlic.com/~lynn/2021.html#26 CMSBACK, ADSM, TSM
HONE posts
https://www.garlic.com/~lynn/subtopic.html#hone
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
--
virtualization experience starting Jan1968, online at home since Mar1970
NSFNET
From: Lynn Wheeler <lynn@garlic.com>
Subject: NSFNET
Date: 20 Aug, 2026
Blog: Facebook
I had taken two credit hour intro to fortran/computers. At the end of
semester, I was hired to rewrite 1401 MPIO in assembler for
360/30. The univ was getting 360/67 for tss/360 replacing 709/1401
(360/30 temporarily replaced 1401 pending arrival of 360/67). The
univ. shutdown datacenter on weekends and I would have the place
dedicated (although 48hrs w/o sleep made monday classes hard). I was
given pile of hardware&software manuals and got to design & implement
monitor, device drivers, interrupt handlers, error recovery, storage
management, etc. and within a few weeks had 2000 card assembler
program.
Within a year of taking intro class, the 360/67 arrives and I was
hired fulltime responsible for OS/360 (tss/360 never came to
production). The 709 ran student fortran jobs in under second, but
initially with os/360, they ran over a minute. I install HASP which
cuts the time in half. I then start redoing SYSGEN STAGE2, to
carefully place datasets and PDS members to optimize disk arm seek and
multi-track search ... cutting another 2/3rds to 12.9secs; it never
got better than 709 until I install UofWaterloo WATFOR, WATFOR clocked
at 20,000 statements/min (333 statements/sec), student fortran tended
to run 30-60 statements.
Then IBM CSC came out to install (virtual machine) CP/67 (3rd
after CSC itself and MIT Lincoln Labs) and I mostly get to play with
it during my weekend 48hr window. I then spend a few months rewriting
pathlengths for running OS/360 in virtual machine. Bare machine
test ran 322secs ... initially 856secs (CP67 CPU 534secs). After a few
months I had CP67 CPU down from 534secs to 113secs. I then start
rewriting the dispatcher/scheduler , (dynamic adaptive resource
manager/default fair share scheduling policy), paging, adding ordered
seek queuing (from FIFO) and mutli-page transfer channel programs
(from FIFO and optimized for transfers/revolution, getting 2301 paging
drum from 70-80 4k transfers/sec to channel transfer peak of 270). Six
months after univ initial CP/67 install, CSC was giving one week class
in LA. I arrive on Sunday afternoon and asked to teach the class, it
turns out that the people that were going to teach it had resigned the
Friday before to join one of the 60s CSC CP67 commercial online
spin-offs.
CP/67 had arrived with 1052&2741 terminal support and
auto-terminal ident, capable switching terminal type scanner type for
each port. Univ. also had TTY33&35 and I add ASCII support
integrated with auto-terminal type. I then want to have single dial-in
number (hunt group) for all terminals. Didn't quite work, IBM had
hard-wired line speed ... so we start a clone terminal
controller. Build a IBM channel interface board for Interdata/3
programmed to emulate IBM controller with the addition for
auto-baud. Then upgraded with Interdata/4 for channel interface and
cluster of Interdata/3s for port interfaces. Interdata (and later
Perkin-Elmer) sell them as clone controllers.
https://en.wikipedia.org/wiki/Interdata
https://en.wikipedia.org/wiki/Perkin-Elmer#Computer_Systems_Division
... and four of us are written up responsible for (some part of) IBM
clone controller business
Turn of century, toured datacenter that had descendant of our
Interdata box, that was handling nearly all payment card POS terminal
dial-up east of the Mississippi.
Before I graduate, I was hired fulltime into very small group in the
Boeing CFO office to help with the formation of Boeing Computer
Services (consolidate all dataprocessing into independent business
unit). I think Renton datacenter was possibly largest in the world,
360/65s arriving faster than they could be installed, boxes constantly
staged in the hallways around the machine room (and lonely 360/75,
black rope around perimeter, black velvet dropped and guards at
perimeter when running classified). I was given mid-management badge
so I could park in lots newer to bldgs. Lots of politics between
Renton director and Boeing CFO, who only had a 360/30 up at Boeing
Field for payroll, although they enlarge the room for a 360/67 for me
to play with when I wasn't doing other stuff. When I graduate, I join
IBM Cambridge Scientific Center (instead of staying w/Boeing CFO).
CSC co-worker was responsible for the Science Center wide-area
network, which morphs into the corporate network (larger than
arpanet/internet from beginning until sometime mid/late 80s (about the
time it was forced to convert to SNA/VTAM) and technology also used
for corporate sponsored univ. BITNET. Account by one of the CSC
inventors of GML in 1969 (decade later morphs into ISO standard SGML
and after another decade morphs into HTML at CERN):
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for
the project, something I was allowed to do part-time because of my
knowledge of the user requirements. My real job was to encourage the
staffs of the various scientific centers to make use of the
CP-67-based Wide Area Network that was centered in Cambridge.
... snip ...
Edson was responsible for CP67 wide-area network
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which
paved the way to the Internet as we know it today.
... snip ...
SJMerc article about Edson (he passed aug2020) and "IBM'S MISSED
OPPORTUNITY WITH THE INTERNET" (gone behind paywall but lives free at
wayback machine)
https://web.archive.org/web/20000124004147/http://www1.sjmercury.com/svtech/columns/gillmor/docs/dg092499.htm
Also from wayback machine, some additional (IBM missed internet)
references from Ed's website
https://web.archive.org/web/20000115185349/http://www.edh.net/bungle.htm
Also early 80s, I got HSDT project, T1 and faster computer links (both
terrestrial and satellite) and battles with SNA group (60s, IBM had
2701 supporting T1, 70s with SNA/VTAM and issues, links were capped at
56kbit ... and I had to mostly resort to non-IBM hardware, did find
the FSD Series/1 T1 Zirpel card for gov. agencies with failing
2701s). Also was working with NSF director and was suppose to get $20M
to interconnect the NSF Supercomputer centers. NSF gives grant to UC
for Berkeley Supercomputer Center and I'm asked to give them
presentations on the interconnect. However, UC Regents plan for next
new bldg was for UCSD and it becomes the San Diego Supercomputer
Center instead, run by General Atomics (NSF didn't have control).
Then congress cuts the budget, some other things happened and
eventually there was RFP released (in part based on what we already
had running). NSF 28Mar1986 Preliminary Announcement:
https://web.archive.org/web/20250428194307/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers
Program to provide Supercomputer cycles; the New Technologies Program
to foster new supercomputer software and hardware developments; and
the Networking Program to build a National Supercomputer Access
Network - NSFnet.
... snip ...
IBM internal politics was not allowing us to bid. The NSF director
tried to help by writing the company a letter (3Apr1986, NSF Director
to IBM Chief Scientist and IBM Senior VP and director of Research,
copying IBM CEO) with support from other gov. agencies ... but that
just made the internal politics worse (as did claims that what we
already had operational was at least 5yrs ahead of the winning bid),
as regional networks connect in, NSFnet becomes the NSFNET backbone,
precursor to modern internet. Note RFP had called for T1 links,
however winning bid put in PC/RTs each with 440kbit/sec link ... then
to make it look something like T1, they put in T1 trunks with telco
multiplexors running multiple 440kbit/sec links over T1 trunks.
Mid-80s, communication group was fighting off mainframe TCP/IP
support. When they lost, the said that since they had corporate
responsibility for everything that crossed datacenter walls, it had to
be released through them. What shipped got aggregate 44kbytes/sec
using nearly whole 3090 CPU. I then add RFC1044 and in some tuning
tests at Cray Research between Cray and 4341, got sustained 4341
channel throughput using only modest amount of 4341 CPU (something
like 500 times improvement in bytes moved per instruction executed).
1988, branch office asks if I could help SLAC/Gustavson with SCI
standard (various uses including shared memory multiprocessor, up to
64 cache consistency, some efforts Data General and Sequent 64 4-i486
boards (256 processors), Convex 64 2-HP-snake boards (128 processor),
SGI 64 4-MIPS boards, etc),
https://en.wikipedia.org/wiki/Scalable_Coherent_Interface
https://www.scizzl.com/
After leaving IBM in the 90s, do some consulting for Steve Chen (had
done YMP at Cray) who was Sequent CTO (before IBM bought it and shut
it down).
I was also on Chessin's XTP Technical Advisory Board and since there
was heavy participation by several gov agencies, we take it as
High-Speed Protocol to ANSI x3s3.3 (ISO chartered standards body) for
level 3&4. Initially saw acceptance but then was told that ISO
required standards work *ONLY* on protocol that conformed to OSI
Model. XTP/HSP didn't because 1) supported internetworking, level that
didn't exist in OSI, 2) skipped level 3/4 interface going directly to
LAN MAC, 3) supported LAN MAC which doesn't exist in OSI.
There was request for NSFNET upgrade to T3. The executive running the
IBM response asked me to be "red team" (possibly thought it would
shutdown my criticism) and 20 people from half dozen labs around the
world, the "blue team". At the review, I presented 1st and then the
"blue team". Ten minutes into the "blue team" presentation, the
executive pounded on the table and said he would lay down in front of
a garbage truck before he would allow my presentation go forward (we
and few others get up and walk out).
Also 1988, IBM branch office asks if I could help LLNL with
standardize some serial stuff they were working with, which quickly
becomes fibre-channel standard ("FCS", including some stuff I had done
in 1980, initially 1gbit transfer, full-duplex, aggregate 200mbyte/sec
transfer). Then IBM POK mainframe releases their serial stuff (when it
is already obsolete), "ESCON", 10mbyte/sec. Then some IBM POK
engineers becomes involved with FCS and define a heavy-weight protocol
that radically reduces throughput eventually released as "FICON".
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
AWD engineer that had done much of the work for RS/6000 "SLA" (sort of
enhanced version of mainframe ESCON, but faster and full-duplex)
... problem wasn't compatible or interoperable with anything
else. Pending availability of "FCS" support, con one of the super high
performance router vendors (16 fast Ethernet, FDDI, T1 & T3,
supercomputer and IBM mainframe channel options) into adding SLA
support. Engineer then wanted to do a 800mbit SLA but talk him into
joining the FCS standards committee instead (part of objective was
RS/6000 as high-performance TCP/IP server ... leveraging the existing
SLA interface to external super high-performance router). Get some
installed in the new Almaden research bldg.
... more 1988, IBM Nick Donofrio approves HA/6000, initially for
NYTimes to move their newspaper system (ATEX) off DEC VAXCluster to
RS/6000, I rename it HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing (technical/scientific) cluster scale-up with
national labs (LANL, LLNL, NCAR/UCAR, etc) and commercial cluster
scale-up with RDBMS vendors (Oracle, Sybase, Ingres, Informix) with
VAXCluster support in same source base with UNIX. Also migrate LLNL
UNICOS "LINCS"
https://pdfs.semanticscholar.org/f5fc/36f760a62a39e1b780b556d5817841d40fac.pdf
and NCAR "Mesa Archive" spin-off, to HA/CMP.
Early Jan92, there was meeting with Oracle CEO and IBM/AWD executive
Hester tells Ellison that we would have 16-system clusters by mid92
and 128-system clusters by ye92. Mid-jan92, I update FSD on HA/CMP
work with national labs and FSD decides to go with HA/CMP for federal
supercomputers. By end of Jan, we are told that cluster scale-up is
being transferred to Kingston for announce as IBM Supercomputer (SP,
technical/scientific *ONLY*) and we aren't allowed to work with
anything that has more than four systems (we leave IBM a few months
later). A couple weeks later, 17feb1992, Computerworld news ... IBM
establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
Sometime after leaving IBM, I had been brought in as consultant into
small client/server startup. Two former Oracle people (that were in
the Ellison/Hester meeting) are there responsible for something they
called "commerce server" and they wanted to do payment
transactions. The startup had also invented this technology they
called "SSL" they wanted to use. I was responsible for everything
between commerce servers (now frequently called e-commerce) and
payment networks. I then do talk: "Why Internet Isn't Business
Critical Dataprocessing" (based on processes, documentation and
software I had to do for e-commerce) that (Internet, IETF) RFC
standards editor Postel sponsored at ISI/USC.
https://en.wikipedia.org/wiki/Jon_Postel
He also had me help with the periodically re-released STD1.
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
RFC1044 posts
https://www.garlic.com/~lynn/subnetwork.html#1044
XTP/HSP posts
https://www.garlic.com/~lynn/subnetwork.html#xtphsp
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
Interop 88 posts
https://www.garlic.com/~lynn/subnetwork.html#interop88
801, risc, romp, iliad, pc/rt, rios, rs/6000, power/pc posts
https://www.garlic.com/~lynn/subtopic.html#801
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
payment gateway posts
https://www.garlic.com/~lynn/subnetwork.html#gateway
--
virtualization experience starting Jan1968, online at home since Mar1970
Tandem
From: Lynn Wheeler <lynn@garlic.com>
Subject: Tandem
Date: 21 Aug, 2026
Blog: Facebook
I transferred out to IBM SJR on the west coast the 2nd half of 70s and
worked with Jim Gray and Vera Watson on the original SQL/Relational,
System/R. Fall of 1980, Jim Gray leaves for Tandem (palming off some
amount of stuff on me).
I was blamed for online computer conferencing late 70s and early
80s. It really took off spring of 1981 when I distributed trip report
of visit to Jim Gray at Tandem. Only about 300 directly participated,
but claims that upwards of 25,000 were reading. Six copies of 300
pages were packaged in Tandem 3-ring binders and sent to the corporate
executive committee (folklore 5of6 wanted to fire me). From IBM Jargon
(Datamation reference was in original version):
https://havantcivicsociety.uk/wp-content/uploads/2019/05/ibmjarg.pdf
Tandem Memos - n. Something constructive but hard to control; a fresh
of breath air (sic). That's another Tandem Memos. A phrase to worry
middle management. It refers to the computer-based conference (widely
distributed in 1981) in which many technical personnel expressed
dissatisfaction with the tools available to them at that time, and
also constructively criticized the way products were [are]
developed. The memos are required reading for anyone with a serious
interest in quality products. If you have not seen the memos, try
reading the November 1981 Datamation summary.
... snip ...
There was executive summary and a summary of the executive summary;
from summary of summary:
• The perception of many technical people in IBM is that the company is
rapidly heading for disaster. Furthermore, people fear that this
movement will not be appreciated until it begins more directly to
affect revenue, at which point recovery may be impossible
• Many technical people are extremely frustrated with their management
and with the way things are going in IBM. To an increasing extent,
people are reacting to this by leaving IBM. Most of the contributors
to the present discussion would prefer to stay with IBM and see the
problems rectified. However, there is increasing skepticism that
correction is possible or likely, given the apparent lack of
commitment by management to take action
• There is a widespread perception that IBM management has failed to
understand how to manage technical people and high-technology
development in an extremely competitive environment
... snip ...
There were task forces to look at online computer conferencing
resulting in official IBM forum software and approved, moderated forum
discussions. Also a researcher was paid to study how I communicated;
sat in the back of my office for nine months taking notes on how I
communicated, also got copies of all my incoming and outgoing email
and logs of all instant messages. The result was used for IBM research
reports, conference talks and papers, books and Stanford Phd (joint
with language and computer AI, Winograd was advisor on computer AI
side).
At Dec81 ACM SIGOPS meeting, Jim asks me if I can help Tandem
co-worker get their Stanford Phd. It involved global LRU page
replacements and the forces from the late 60s ACM local LRU are
lobbying to block giving Phd for anything involving global
LRU. I had lots of data on my 60s undergraduate Global LRU
work and at CSC, that run 768kbyte 360/67 (104 pageable pages after
fixed requirement) with 75-80 users. I also had lots of data from the
IBM Grenoble Science Center that had modified CP67 to conform to the
60s ACM local LRU literature (1mbyte 360/67, 155 pageable pages after
fixed requirement). CSC with 75-80 users had better response and
throughput (104 pages) than Grenoble running 35 users (similar
workloads and 155pages). For some reason, IBM executives (also)
prohibited me from responding for nearly a year (19Oct1982).
copy of old archived post with 19Oct1982 cover letter
https://web.archive.org/web/20260311153642/https://www.garlic.com/~lynn/2006w.html#email821019
original SQL/Relational, System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
online computer conferencing posts
https://www.garlic.com/~lynn/subnetwork.html#cmc
global LRU, paging posts
https://www.garlic.com/~lynn/subtopic.html#wsclock
--
virtualization experience starting Jan1968, online at home since Mar1970
MVT Storage Management Problems
From: Lynn Wheeler <lynn@garlic.com>
Subject: MVT Storage Management Problems
Date: 22 Aug, 2026
Blog: Facebook
Early last decade I was asked to track down decision to add virtual
memory to all 370s. Basically MVT storage management was so bad that
region sizes had to be specified four times larger than used, limited
standard 1mbyte, 370/165 to four concurrent running regions,
insufficient to keep system busy and justified.
Running MVT in 16mbyte virtual address space (similar to running MVT
in a 360/67, CP/67 16mbyte virtual machine) allowed number of
concurrent regions to be increased by factor of four times (capped at
15 concurrent regions because of 4bit storage protect key) with little
or no paging.
Ludlow was doing the initial implementation of MVT->VS2/SVS on
360/67 (until engineering 370 with virtual memory) and I would drop by
periodically. He had a little bit of code for the 16mbyte virtual
address space and some simple paging. Biggest task was channel
programs passed to EXCP/SVC0 now had virtual addresses and channels
required real addresses and he borrows CP67's CCWTRANS for integrating
into EXCP (creating channel program copies, replacing virtual with
real addresses).
VS2/SVS->VS2/MVS eliminated storage protection keys to separate
"regions" and the 15 region concurrent cap ... placing them in
separate virtual address spaces. But that created a different
problem. OS/360 history was heavily built on pointer passing API. As a
result an 8mbyte "image" of MVS kernel was placed in every 16mbyte
virtual address space (leaving 8mbyte for applications). Then because
subsystems were placed in their own 16mbyte virtual address
space. Then for applications to call subsystems, they created the
1mbyte common segment area ("CSA") image in every 16mbyte virtual
address space (for passing data to and back), leaving 7mbytes for
applications. CSA space needs were proportional to concurrently
running applications and subsystems and by 3033 "CSA" exploded to 5-6
mbytes (and rename "common system area") ... leaving 2-3mbytes for
applications ... but CSA threatening to become 8mbytes (leaving zero
space for applications) .... part of desperate need to move customer
base to MVS/XA (also the retrofitting "dual-address" mode facility to
3033).
Before I graduate in the 60s, I'm hired fulltime into small group in
Boeing CFO office to help with formation of Boeing Computer Services
(consolidate all dataprocessing into independent business unit). I
think Renton, largest datacenter in the world, 360/65s arriving faster
than they could be installed, boxes constantly staged in hallways
around machine room (did have lone 360/75, when doing classified work;
black rope around area, guards at corners, heavy black felt covering
console lights and visible 1403 printer areas). Lots of politics
between Renton director and CFO, who only had a 360/30 up at Boeing
Field for payroll, although they enlarge the room to install 360/67
for me to play with when I wasn't doing other stuff. I had mid-level
management badge so could park in lots closer to the bldgs. 747-3 was
flying skies of Seattle getting FAA flt certification. Tours of
mock-up of 747 cabin just south of Boeing field would claim 747s
carried so many passengers, they would never be serviced with fewer
than four jetways. Both Boeing and IBM told stories that on 360
announcement day, Boeing submitted an order making the IBM marketing
rep, highest paid IBM employee that year. Sales were still straight
commission; IBM shifts to quota the following year.
Boeing Huntsville ordered a 2-CPU 360/67 with lots of 2250 graphic
displays for TSS/360 ... TSS/360 didn't come to production so they
configured it as two 360/65 MVT13 systems. They ran into the MVT
storage management problems (later motivating IBM adding virtual
memory to all 370s and MVT morphing into VS2) and modified MVT13 to
run in simplified virtual memory mode as countermeasure
Posts mentioning VS2, Ludlow, MVS, Common Segment Area, 3033
https://www.garlic.com/~lynn/2026b.html#70 IBM Online Systems
https://www.garlic.com/~lynn/2026.html#12 IBM Virtual Machine and Virtual Memory
https://www.garlic.com/~lynn/2025e.html#115 Comsat
https://www.garlic.com/~lynn/2025e.html#105 Early Mainframe Work
https://www.garlic.com/~lynn/2025e.html#71 IBM 370 Virtual Memory
https://www.garlic.com/~lynn/2025c.html#108 IBM OS/360
https://www.garlic.com/~lynn/2025c.html#79 IBM System/360
https://www.garlic.com/~lynn/2024d.html#24 ARM is sort of channeling the IBM 360
https://www.garlic.com/~lynn/2023g.html#77 MVT, MVS, MVS/XA & Posix support
https://www.garlic.com/~lynn/2023f.html#26 Ferranti Atlas
https://www.garlic.com/~lynn/2022d.html#55 CMS OS/360 Simulation
https://www.garlic.com/~lynn/2016.html#78 Mainframe Virtual Memory
Posts mentioning Boeing Huntsville, 360/67, 2250s, MVT13 and MVT storgage
management problems
https://www.garlic.com/~lynn/2026.html#59 IBM CP67 and VM370
https://www.garlic.com/~lynn/2026.html#48 UofM MTS and IBM CP67
https://www.garlic.com/~lynn/2025b.html#117 SHARE, MVT, MVS, TSO
https://www.garlic.com/~lynn/2024g.html#72 Building the System/360 Mainframe Nearly Destroyed IBM
https://www.garlic.com/~lynn/2024e.html#136 HASP, JES2, NJE, VNET/RSCS
https://www.garlic.com/~lynn/2024d.html#40 ancient OS history, ARM is sort of channeling the IBM 360
https://www.garlic.com/~lynn/2024b.html#3 Bypassing VM
https://www.garlic.com/~lynn/2024.html#87 IBM 360
https://www.garlic.com/~lynn/2023g.html#81 MVT, MVS, MVS/XA & Posix support
https://www.garlic.com/~lynn/2023g.html#39 Vintage Mainframe
https://www.garlic.com/~lynn/2023g.html#19 OS/360 Bloat
https://www.garlic.com/~lynn/2023g.html#5 Vintage Future System
https://www.garlic.com/~lynn/2023f.html#110 CSC, HONE, 23Jun69 Unbundling, Future System
https://www.garlic.com/~lynn/2023f.html#52 IBM Vintage 1130
https://www.garlic.com/~lynn/2023e.html#34 IBM 360/67
https://www.garlic.com/~lynn/2021k.html#106 IBM Future System
https://www.garlic.com/~lynn/2010c.html#4 Processes' memory
https://www.garlic.com/~lynn/2010b.html#61 Source code for s/360 [PUBLIC]
https://www.garlic.com/~lynn/2009r.html#43 Boeings New Dreamliner Ready For Maiden Voyage
https://www.garlic.com/~lynn/2007m.html#60 Scholars needed to build a computer history bibliography
https://www.garlic.com/~lynn/2004c.html#47 IBM 360 memory
https://www.garlic.com/~lynn/2001m.html#55 TSS/360
--
virtualization experience starting Jan1968, online at home since Mar1970
MVS
From: Lynn Wheeler <lynn@garlic.com>
Subject: MVS
Date: 25 Aug, 2026
Blog: Facebook
I had taken two credit hr intro to Fortran/Computers. At the end of
semester was hired to rewrite 1401 MPIO in 360 assembler for 360/30.
Univ was getting 360/67 for tss/360 replacing 709/1401 and got 360/30
temporarily replacing 1401 pending arrival of 360/67. Univ. 360/67
arrived within year of taking intro class and I was hired fulltime
responsible for OS/360 (TSS/360 not coming to production, so ran as
360/65). 709 did student fortran in less than second. 360/67 (as
360/65) os/360 (MFT9.5) took over a minute, I install HASP cutting
time in half. MFT11, I start redoing SYSGEN STAGE2 to carefully place
datasets and PDS members to optimize arm seek and (PDS directory)
multi-track search, cutting another 2/3rds to 12.9secs. OS/360 never
got better than 709 until I install Univ Waterloo WATFOR; as 360/65,
WATFOR clocked at 20,000 statements/min (333 statements/sec), student
fortran tended to run 30-60 statements.
Before I graduate, I'm hired fulltime into small group in the Boeing
CFO office (to help with the formation of Boeing Computer Services,
consolidate all dataprocessing into independent business unit). I
think Renton datacenter largest in the world. Boeing & IBMers tell
story on day 360 was announced, Boeing walks into the IBM marketing
rep's office and gives him an order making him the highest paid
employee that year (days of straight commission, following year, IBM
deploys quota).
After I graduate, I join IBM Cambridge Science Center (instead of
staying with Boeing CFO) and one of my hobbies was enhanced production
operating systems for internal datacenters and sales&marketing
support HONE systems were one of my first (and long time) customers.
After Future System imploded in mid-70
http://www.jfsowa.com/computer/memo125.htm
https://en.wikipedia.org/wiki/IBM_Future_Systems_project
https://people.computing.clemson.edu/~mark/fs.html
I got coned into helping with a 16-CPU 370 multiprocessor effort and
we talk the 3033 processor engineers into working on it in their spare
time (a lot more interesting than mapping 168 logic to 20% faster
chips). Everybody thought it was great until somebody tells the head
of POK that it could be decades before POK's favorite son operating
system ("MVS") has (effective) 16-CPU support (POK doesn't ship 16-CPU
until after turn of century; MVS documents at the time was that MVS
2-CPU support only got 1.2-1.5 throughput of single processor
operation). Then some of us were invited to never visit POK again and
the 3033 processor engineers were directed, heads down and no
distractions (once 3033 is out the door, the 3033 processor engineers
start on trout/3090).
After being forbidden to visit POK again (although 3033 proceessor
engineers would have me sneak back in), I transfer to SJR on the west
coast get to wander around datacenters in silicon valley, including
disk bldg14 (engineering) and bldg15 (product test) across the
street. They were doing 7x24, prescheduled, stand-alone testing
(mentioned that they had recently tried "MVS", but it had 15min MTBF
in that environment, requiring manual re-ipl). I offer to rewrite I/O
supervisor making it bullet proof and never fail; supporting any
amount of on-demand concurrent testing (greatly improving
productivity). I write an (internal only) I/O Reliability Enhancement research
report and happen to mention MVS 15min MTBF bringing down the wrath of
the MVS organization on my head (perceived that they were concerned
about IBM executives would see it).
Then bldg15 get 1st engineering 3033 outside POK processor
engineering. Product test was only taking percent or two of CPU, so we
scrounge up a 3830 controller and 3330 string, setting up our own
private online service (I would frequently have dedicated for days at
a time).
1988, branch office asks if I could help LLNL with standardize some
serial stuff they were working with, which quickly becomes
fibre-channel standard ("FCS", including some stuff I had done in
1980, initially 1gbit transfer, full-duplex, aggregate 200mbyte/sec
transfer). Then IBM POK mainframe releases their serial stuff (when it
is already obsolete), "ESCON", 10mbyte/sec. Then some IBM POK
engineers becomes involved with FCS and define a heavy-weight protocol
that radically reduces throughput eventually released as "FICON".
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
Also 1988, Nick Donofrio approves HA/6000, originally for NYTimes to
move their newspaper system ("ATEX") off DEC VAXCluster to RS/6000
(running project out at Los Gatos lab). I rename it HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing technical/scientific cluster scale-up with national
labs (LANL, LLNL, NCAR, etc) and commercial cluster scale-up with
RDBMS vendors (Oracle, Sybase, Ingres, Informix) with VAXCluster
support in same source base with Unix. I was planning on using FCS for
HA/CMP FCS and ported LLNL's UNICOS LINCS and NCAR's filesystem (spun
off as Mesa Archival) to HA/CMP.
IBM S/88 (relogo'ed Stratus) Product Administrator started taking us
around to their customers and also had me write a section for the
corporate continuous availability strategy document (it gets pulled
when both AS400/Rochester and mainframe/POK complain they couldn't
meet requirements). Had coined disaster survivability and
geographic survivability (as counter to disaster/recovery) when out
marketing HA/CMP. One of the visits to 1-800 bellcore development
showed that S/88 would use a century of downtime in one software
upgrade, while HA/CMP had a couple extra "nines" (compared to S/88).
One of the first HA/CMP customer installs was new Indian Reservation
Casino in Connecticut, was suppose to have week of testing before
opening ... but after 24hrs, they decided to open the doors (based on
projected revenue; at the time was largest in the US, still one of the
largest in the country)
https://en.wikipedia.org/wiki/Foxwoods_Resort_Casino#Debt_default
Early Jan92, there was HA/CMP meeting with Oracle CEO and IBM/AWD
executive Hester tells Ellison that we would have 16-system clusters
by mid92 and 128-system clusters by ye92. Mid-jan92, I update FSD on
HA/CMP work with national labs and FSD decides to go with HA/CMP for
federal supercomputers. By end of Jan92, we are told that cluster
scale-up is being transferred to Kingston for announce as IBM
Supercomputer (SP, technical/scientific *ONLY*) and we aren't allowed
to work with anything that has more than four systems (we leave IBM a
few months later). A couple weeks later, 17feb1992, Computerworld news
... IBM establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
Some speculation that commercial HA/CMP would have eaten IBM mainframe
in commercial market ... 1993 benchmark (number of program iterations
compared to industry MIPS reference platform):
ES/9000-982 : 8CPU 408MIPS, (51MIPS/CPU)
RS6000/990 (RIOS chipset) : 1-CPU: 126MIPS
I then asked in as consultant to small client/server startup. Two
former Oracle people (that were in the Ellison/Hester meeting) are
there responsible for something they called "commerce server" and they
wanted to do payment transactions. The startup had also invented this
technology they called "SSL" they wanted to use. I was responsible for
everything between commerce servers (now frequently called e-commerce)
and payment networks. I then do talk: "Why Internet Isn't Business
Critical Dataprocessing" (based on processes, documentation and
software I had to do for e-commerce) that (Internet, IETF) RFC
standards editor Postel sponsored at ISI/USC.
https://en.wikipedia.org/wiki/Jon_Postel
He also had me help with the periodically re-released STD1.
HASP, JES2, ASP, JES3, NJI, NJE posts
https://www.garlic.com/~lynn/submain.html#hasp
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
SMP, tightly-coupled, shared memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
getting to play disk engineer posts
https://www.garlic.com
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
e-commerce gateway posts
https://www.garlic.com/~lynn/subnetwork.html#gateway
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
channel-extender posts
https://www.garlic.com/~lynn/submisc.html#channel.extender
FCS and/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
posts mentioning CKD, DASD, multi-track search, FBA
https://www.garlic.com/~lynn/submain.html#dasd
--
virtualization experience starting Jan1968, online at home since Mar1970
MVS
From: Lynn Wheeler <lynn@garlic.com>
Subject: MVS
Date: 25 Aug, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#25 MVS
Early 80s, got HSDT, T1 and faster computer links (both terrestrial
and satellite) and battles with the communication group (60s, IBM had
2701 supporting T1, 70s with SNA/VTAM and issues, links were capped at
56kbit ... and I had to mostly resort to non-IBM hardware, did find
the FSD Series/1 T1 Zirpel card for gov. agencies with failing
2701s). First cross country T1 was between Los Gatos lab and
Clementi's
https://en.wikipedia.org/wiki/Enrico_Clementi
E&S lab in Kingston that had a whole boatload of FPS-264 systems.
https://en.wikipedia.org/wiki/Floating_Point_Systems
that had 40mbyte/sec disk arrays.
Communication group was fighting off release of mainframe TCP/IP
support. When they lost they said it had to be released through them
since they had corporate ownership of everything that crossed
datacenter walls. What shipped got aggregate 44kbytes/sec using nearly
whole 3090 processor (also shipped on MVS by simulating VM370
"diagnose" instructions). I then added RFC1044 support and in some
tuning tests at Cray Research, between Cray and 4341, got sustained
4341 channel throughput using only modest amount of 4341 CPU
(something like 500 times improvement in bytes moved per instruction
executed).
Also was working with NSF director and was suppose to get $20M to
interconnect the NSF Supercomputer centers. NSF gives grant to UC for
Berkeley Supercomputer Center and I'm asked to give them presentations
on the interconnect. However, UC Regents plan for next new bldg was
for UCSD and it becomes the San Diego Supercomputer Center instead,
run by General Atomics (NSF didn't have control). Then congress cuts
the budget, some other things happened and eventually there was RFP
released (in part based on what we already had running). NSF 28Mar1986
Preliminary Announcement:
https://web.archive.org/web/20251217225848/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers
Program to provide Supercomputer cycles; the New Technologies Program
to foster new supercomputer software and hardware developments; and
the Networking Program to build a National Supercomputer Access
Network - NSFnet.
... snip ...
IBM internal politics was not allowing us to bid. The NSF director
tried to help by writing the company a letter (3Apr1986, NSF Director
to IBM Chief Scientist and IBM Senior VP and director of Research,
copying IBM CEO) with support from other gov. agencies ... but that
just made the internal politics worse (as did claims that what we
already had operational was at least 5yrs ahead of the winning bid),
as regional networks connect in, NSFnet becomes the NSFNET backbone,
precursor to modern internet. Note RFP had called for T1 links,
however winning bid put in PC/RTs each with 440kbit/sec link ... then
to make it look something like T1, they put in T1 trunks with telco
multiplexors running multiple 440kbit/sec links over T1 trunks.
Early 90s, communication group hired a silicon valley contractor
directly implement TCP/IP in VTAM. What he demo'ed showed TCP running
much faster than LU6.2. He was then told that everybody "knows" that a
"proper" TCP/IP implementation is much slower than LU6.2 ... and they
would only be paying for a "proper" implementation.
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
RFC1044 posts
https://www.garlic.com/~lynn/subnetwork.html#1044
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
a few posts mentioning Clementi, FPS-264 and 40mbyte/sec disk arrays
https://www.garlic.com/~lynn/2025c.html#116 Internet
https://www.garlic.com/~lynn/2025c.html#107 IBM San Jose Disk
https://www.garlic.com/~lynn/2025b.html#74 Cluster Supercomputing
https://www.garlic.com/~lynn/2025b.html#37 FAA ATC, The Brawl in IBM 1964
https://www.garlic.com/~lynn/2025.html#114 IBM 370 Virtual Memory
https://www.garlic.com/~lynn/2025.html#54 Multics vs Unix
https://www.garlic.com/~lynn/2025.html#33 IBM ATM Protocol?
https://www.garlic.com/~lynn/2025.html#31 On-demand Supercomputer
https://www.garlic.com/~lynn/2025.html#12 IBM APPN
https://www.garlic.com/~lynn/2024f.html#82 IBM Registered Confidential and "811"
https://www.garlic.com/~lynn/2024e.html#116 what's a mainframe, was is Vax addressing sane today
https://www.garlic.com/~lynn/2024e.html#79 NSFNET
https://www.garlic.com/~lynn/2024e.html#28 VMNETMAP
https://www.garlic.com/~lynn/2024d.html#85 ATT/SUN and Open System Foundation
https://www.garlic.com/~lynn/2024d.html#19 IBM Internal Network
https://www.garlic.com/~lynn/2024d.html#5 Disconnect Between Coursework And Real-World Computers
https://www.garlic.com/~lynn/2024c.html#68 Berkeley 10M
https://www.garlic.com/~lynn/2024b.html#22 HA/CMP
https://www.garlic.com/~lynn/2024.html#38 RS/6000 Mainframe
https://www.garlic.com/~lynn/2023g.html#97 Shared Memory Feature
https://www.garlic.com/~lynn/2023f.html#44 IBM Vintage Series/1
https://www.garlic.com/~lynn/2023f.html#27 Ferranti Atlas
https://www.garlic.com/~lynn/2023d.html#104 DASD, Channel and I/O long winded trivia
https://www.garlic.com/~lynn/2023d.html#3 IBM Supercomputer
https://www.garlic.com/~lynn/2023b.html#57 Ethernet (& CAT5)
https://www.garlic.com/~lynn/2022h.html#26 Inventing the Internet
https://www.garlic.com/~lynn/2022c.html#57 ASCI White
https://www.garlic.com/~lynn/2022c.html#52 IBM Personal Computing
https://www.garlic.com/~lynn/2022c.html#22 Telum & z16
https://www.garlic.com/~lynn/2022b.html#79 Channel I/O
https://www.garlic.com/~lynn/2022b.html#69 ARPANET pioneer Jack Haverty says the internet was never finished
https://www.garlic.com/~lynn/2022b.html#16 Channel I/O
https://www.garlic.com/~lynn/2022.html#121 HSDT & Clementi's Kingston E&S lab
https://www.garlic.com/~lynn/2022.html#95 Latency and Throughput
https://www.garlic.com/~lynn/2021j.html#31 IBM Downturn
https://www.garlic.com/~lynn/2021c.html#53 IBM CEO
https://www.garlic.com/~lynn/2021b.html#1 Will The Cloud Take Down The Mainframe?
https://www.garlic.com/~lynn/2021.html#62 Mainframe IPL
https://www.garlic.com/~lynn/2019c.html#48 IBM NUMBERS BIPOLAR'S DAYS WITH G5 CMOS MAINFRAMES
https://www.garlic.com/~lynn/2019.html#32 Cluster Systems
https://www.garlic.com/~lynn/2018f.html#109 IBM Token-Ring
https://www.garlic.com/~lynn/2018b.html#47 Think you know web browsers? Take this quiz and prove it
https://www.garlic.com/~lynn/2017j.html#92 mainframe fortran, or A Computer That Never Was: the IBM 7095
https://www.garlic.com/~lynn/2017h.html#50 System/360--detailed engineering description (AFIPS 1964)
https://www.garlic.com/~lynn/2014b.html#4 IBM Plans Big Spending for the Cloud ($1.2B)
--
virtualization experience starting Jan1968, online at home since Mar1970
HSDT
From: Lynn Wheeler <lynn@garlic.com>
Subject: HSDT
Date: 26 Aug, 2026
Blog: Facebook
1980, STL was bursting at the seams and 300 people (& 3270
terminals) were being moved to offsite bldg with dataprocessing
service back to STL datacenter. They had tried remote 3270 support,
but found the human factors unacceptable. I then get con'ed into doing
channel-extender support so channel attached 3270 controllers were
placed at offsite bldg with no perceptible difference in human
factors.
STL had been spreading the channel-attached controllers on all
channels with 3830 disk controllers. Unattended side-effect of the
channel extender was it significantly reduced the channel busy
(interference with 3330 DASD) for same amount of 3270 terminal traffic
... increasing system throughput by 10-15%. There was then some
consideration of moving ALL 3270 channel attached controllers to
channel-extenders.
Then did something similar for a IBM FE group in Boulder being moved
to a bldg on the other side of highway. Had infrared modems on the
roofs of the two bldgs. There was complaints about weather blocking
transmissions ... but bit-error tester showed worst case was white-out
snow storm interference when people weren't able to get into the
office. The actual worst case was strong sunlight heated the bldg on
one side, tilting it just enough that the infrared beam got out of
alignment. Had to do some careful positioning to counteract strong sun
light from misaligned the signal.
Early 80s got HSDT, T1 and faster computer links (both satellite and
terrestrial) and battles with communication group (60s, IBM had 2701
that supported T1, issues with SNA in the 70s, capped controllers at
56kbit).
Then my HSDT project got custom built 3-node Ku-band satellite system
(4.5m dishes in Los Gatos and Yorktown and 7m dish in Austin) with
transponder on SBS4. Yorktown community meeting complained about
radiation from the 25watt signal. It was pointed out if they were hung
directly above dish transmission, it would be significantly less
radiation than they were currently getting from local FM radio tower
transmission.
As undergraduate at the univ. in the 60s, I did computer dynamic
adaptive resource scheduling ... so for HSDT, did dynamic adaptive
rate-based pacing ... adapting to both rate and distance/latency
(single hop sat was twice 22k miles, to/back four times. double hop
was four time 22kmiles, round-trip eight times; west/east coast, east
coast/Europe).
Also in 60s, CP67 installed at the univ (3rd after CSC & MIT
Lincoln Labs) supported 1052/2471 telecommunication unit dynamically
swapping terminal-type port scanner. Univ. had some number of
ASCII/TTY and to ASCII-type port scanner (arrived in Heathkit box). I
added ASCII/TTY terminal support. I then wanted to have single dial-in
number ("hunt group") but IBM had taken short-cut and hard-wired each
port line-speed ... so Univ starts a clone terminal controller
project. Build a IBM channel interface board for Interdata/3
programmed to emulate IBM controller with the addition for
auto-baud. Then upgraded with Interdata/4 for channel interface and
cluster of Interdata/3s for port interfaces. Interdata (and later
Perkin-Elmer) sell them as clone controllers.
https://en.wikipedia.org/wiki/Interdata
https://en.wikipedia.org/wiki/Perkin-Elmer#Computer_Systems_Division
... and four of us are written up responsible for (some part of) IBM
clone controller business
Turn of century, toured datacenter that had descendant of our
Interdata box, that was handling nearly all payment card POS terminal
dial-up east of the Mississippi.
Co-worker at CSC (I joined after graduating) was responsible for
RSCS/VNET ... originally for CP67-based wide-area science center
network (which later morphs into corporate internal network and
technology also used for the corporate sponsored Univ BITNET). Ref by
one of the CSC inventors of GML in 1969 (decade later morphs into SGML
and after another decade morphs into HTML at CERN).
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for
the project, something I was allowed to do part-time because of my
knowledge of the user requirements. My real job was to encourage the
staffs of the various scientific centers to make use of the
CP-67-based Wide Area Network that was centered in Cambridge.
... snip ...
Edson was responsible for CP67 wide-area network
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which
paved the way to the Internet as we know it today.
... snip ...
Channel Extender posts
https://www.garlic.com/~lynn/submisc.html#channel.extendeer
Dynamic adaptive resource scheduling posts
https://www.garlic.com/~lynn/subtopic.html#fairshare
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
Clone (plug-compatible) controller posts
https://www.garlic.com/~lynn/submain.html#360pcm
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
BITNET (&/or EARN) posts
https://www.garlic.com/~lynn/subnetwork.html#bitnet
GML, SGML, HTML posts
https://www.garlic.com/~lynn/submain.html#sgml
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
--
virtualization experience starting Jan1968, online at home since Mar1970
HSDT
From: Lynn Wheeler <lynn@garlic.com>
Subject: HSDT
Date: 26 Aug, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#27 HSDT
more HSDT & other trivia:
Early 80s, was having some special designed, custom hardware being
built on the other side of Pacific. Friday (before visit to see
progress), Raleigh distributes an announcement of new
telecommunication, online FORUM with the following definitions
low-speed: 9.6kbits/sec
medium-speed: 19.2kbits/sec
high-speed: 56kbits/sec
very high-speed: 1.5mbits/sec
Monday morning on conference room wall on the other side of Pacific:
low-speed: <20mbits/sec
medium-speed: 100mbits/sec
high-speed: 200mbits-300mbits/sec
very high-speed: >600mbits/sec
... snip ...
trivia: In late 70s and early 80s, I had been blamed for online
computer conferencing on the internal network. It really took off
spring of 1981 when I distributed trip report of visit to Jim Gray at
Tandem. Only about 300 directly participated, but claims that upwards
of 25,000 were reading. Six copies of 300 pages were packaged in
Tandem 3-ring binders and sent to the corporate executive committee
(folklore: 5of6 wanted to fire me). From IBM Jargon (Datamation
reference was in original version):
https://havantcivicsociety.uk/wp-content/uploads/2019/05/ibmjarg.pdf
Tandem Memos - n. Something constructive but hard to control; a fresh
of breath air (sic). That's another Tandem Memos. A phrase to worry
middle management. It refers to the computer-based conference (widely
distributed in 1981) in which many technical personnel expressed
dissatisfaction with the tools available to them at that time, and
also constructively criticized the way products were [are]
developed. The memos are required reading for anyone with a serious
interest in quality products. If you have not seen the memos, try
reading the November 1981 Datamation summary.
... snip ..
There was executive summary and a summary of the executive summary;
from summary of summary:
• The perception of many technical people in IBM is that the company is
rapidly heading for disaster. Furthermore, people fear that this
movement will not be appreciated until it begins more directly to
affect revenue, at which point recovery may be impossible
• Many technical people are extremely frustrated with their management
and with the way things are going in IBM. To an increasing extent,
people are reacting to this by leaving IBM. Most of the contributors
to the present discussion would prefer to stay with IBM and see the
problems rectified. However, there is increasing skepticism that
correction is possible or likely, given the apparent lack of
commitment by management to take action
• There is a widespread perception that IBM management has failed to
understand how to manage technical people and high-technology
development in an extremely competitive environment
... snip ...
There were task forces to look at online computer conferencing
resulting in official IBM FORUM software and approved, moderated FORUM
discussions. Also a researcher was paid to study how I communicated;
sat in the back of my office for nine months taking notes on how I
communicated (face-to-face, telephone), also got copies of all my
incoming and outgoing email and logs of all instant messages. The
result was used for IBM research reports, conference talks and papers,
books and Stanford Phd (joint with language and computer AI, Winograd
was advisor on computer AI side).
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
online computer conferencing posts
https://www.garlic.com/~lynn/subnetwork.html#cmc
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
--
virtualization experience starting Jan1968, online at home since Mar1970
HSDT
From: Lynn Wheeler <lynn@garlic.com>
Subject: HSDT
Date: 27 Aug, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#28 HSDT
IRMA boards should have been connected via COAX to IBM (channel
attached) 3274 controllers ... and it would be problem with IBM 3274
driving 100% channel busy?
https://en.wikipedia.org/wiki/Irma_board
fix would be 3174 ... handling coax & lans?
https://en.wikipedia.org/wiki/IBM_3270#3174_controller
trivia: For PC/RT (with AT bus), AWD did their own cards including
4mbit token-ring. Then for RS/6000 with microchannel, they were told
they couldn't do their own cards but had to use the (communication
group heavily performance kneecapped) PS2 microchannel cards. The
PC/RT 4mbit T/R card had higher throughput than PS2 16mbit T/R
(microchannel) card.
Mid-80s, communication group was fighting release of mainframe TCP/IP
support. When they lost, they said that since they have corporate
responsibility for everything that crossed datacenter walls, it had to
be released through them. What shipped got aggregate 44kbytes/sec
throughput using nearly full 3090 processor. I then add RFC1044
support and in some tuning tests at Cray Research between Cray and
4341, got sustained 4341 channel throughput using only modest amount
of 4341 processor (something like 500 times improvement in bytes moved
per instruction executed).
New Almaden bldg was heavily provisioned with IBM (twisted pair)
wiring presuming 16mbit T/R ... but found 10mbit Ethernet had lower
latency and higher aggregate LAN throughput (than 16mbit T/R). Also
$69 10mbit Ethernet cards had higher card throughput than $800 PS2
(microchannel) 16mbit token-ring cards.
1988, branch office asks if I could help LLNL with standardize some
serial stuff they were working with, which quickly becomes
fibre-channel standard ("FCS", including some stuff I had done in
1980, initially 1gbit transfer, full-duplex, aggregate 200mbyte/sec
transfer). Then IBM POK mainframe releases their serial stuff (when it
is already obsolete), "ESCON", 10mbyte/sec.
Then some IBM POK engineers becomes involved with FCS and define a
heavy-weight protocol that radically reduces throughput eventually
released as "FICON". Latest public benchmark I've found is 2010 z196
"Peak I/O" getting 2M IOPS using 104 FICON (20K IOPS/FICON). About the
same time a native FCS is announced for E5-2600 server blades getting
over million IOPS (two such FCS higher throughput than 104 FICON
running over FCS). Note IBM docs has SAPs (system assist processors
that do actual I/O) CPU be kept to 70% ... or 1.5M IOPS ... also no
CKD DASD have been made for decades (just simulated with industry
standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
AWD engineer that had done much of the work for RS/6000 "SLA" (sort of
enhanced version of mainframe ESCON, but faster and full-duplex, more
like aggregate >40mbytes/sec) ... problem wasn't compatible or
interoperable with anything else. Pending availability of "FCS"
support, con one of the super high performance router vendors (16 fast
Ethernet, FDDI, T1 & T3, supercomputer and IBM mainframe channel
options) into adding SLA support. Engineer then wanted to do a 800mbit
SLA but talk him into joining the FCS standards committee instead
(part of objective was RS/6000 as high-performance TCP/IP server
... leveraging the existing SLA interface to external super
high-performance router). Get some installed in the new Almaden
research bldg
801/risc, iliad, romp, rios, pc/rt, rs/6000, power, power/pc posts
https://www.garlic.com/~lynn/subtopic.html#801
RFC1044 posts
https://www.garlic.com/~lynn/subnetwork.html#1044
FCS &/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
--
virtualization experience starting Jan1968, online at home since Mar1970
SQL Server
From: Lynn Wheeler <lynn@garlic.com>
Subject: SQL Server
Date: 29 Aug, 2026
Blog: Facebook
I transferred out to IBM SJR on the west coast the 2nd half of 70s and
worked with Jim Gray and Vera Watson on the original SQL/Relational,
System/R. Fall of 1980, Jim Gray leaves for Tandem (palming off some
amount of stuff on me). Opposition from the next great DBMS, EAGLE
product possibly accounted for difficulty shipping RDBMS. While the
company was preoccupied with EAGLE, was able to do tech transfer to
Endicott for SQL/DS product. Then when EAGLE imploded, there was
request for how fast could System/R be ported to "MVS", which was
eventually released as "DB2", originally for "decision support" only.
HA/6000 was approved 1988, originally for NYTimes to move their
newspaper system ("ATEX") from DEC VAXCluster to RS/6000. I rename it
HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing technical/scientific cluster scale-up with national
labs (LANL, LLNL, NCAR, etc) and commercial cluster scale-up with
RDBMS vendors (Oracle, Sybase, Ingres, Informix) with VAXCluster
support in same source base with Unix.
Spent some amount of time with Epstein (at Sybase). Epstein had left
the UCB Relational Ingres for Britain-Lee as Chief Scientist ... which
was doing fast DBMS disk controller (offloading much of DBMS
processing to the controller). One of their biggest customers was
large government agency that was also a long-time CP67 (and then
VM370) customer. When Epstein leaves for Teradata, Britain-Lee was
trying to hire Epstein's replacement from the System/R group and the
person that agreed to go, tried to strong-arm me into going with
him. Note Epstein later leaves Teradata to found Sybase (would license
Sybase to Microsoft for SQL Server).
IBM Toronto lab was working on "Shelby" RDBMS for OS2 (but had few
production features, portability, cluster scaleup, etc that were
needed for HA/CMP) ... eventually renamed DB2 also.
Early Jan1992 had meeting with Oracle CEO, where AWD Hester tells
Ellison that we would have 16-system (RS/6000) clusters by mid92 and
128-system clusters by ye92. Late, Jan1992 was told that cluster
scaleup was being moved to Kingston (for announce as SP, IBM
Supercomputer, technical/scientific only) and we weren't allowed to
work on anything with more than four systems in cluster (we leave IBM
a few months later). A couple weeks later, 17feb1992, Computerworld
news ... IBM establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
Some speculation that commercial HA/CMP would have eaten IBM mainframe
in commercial market ... 1993 benchmark (number of program iterations
compared to industry MIPS reference platform):
ES/9000-982 : 8CPU 408MIPS, (51MIPS/CPU)
RS6000/990 (RIOS chipset) : 1-CPU: 126MIPS
trivia: Jim left Tandem (& DEC) and eventually shows up as head of
m'soft San Francisco research lab ... he then cons me into
interviewing in Redmond for chief security architect ... but we
couldn't agree on what needed to be done.
System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
a few old "SQL Server" posts
https://www.garlic.com/~lynn/2023e.html#42 Systems Network Architecture
https://www.garlic.com/~lynn/2021b.html#29 IBM Recruiting
https://www.garlic.com/~lynn/2016f.html#2 IBM DASD RAS discussion
https://www.garlic.com/~lynn/2013o.html#44 the suckage of MS-DOS, was Re: 'Free Unix!
https://www.garlic.com/~lynn/2010g.html#33 SQL Server replacement
https://www.garlic.com/~lynn/2000e.html#49 How did Oracle get started?
--
virtualization experience starting Jan1968, online at home since Mar1970
Punch Cards
From: Lynn Wheeler <lynn@garlic.com>
Subject: Punch Cards
Date: 01 Sep, 2026
Blog: Facebook
I was taking a 2hr intro to fortran/computers, student keypunch room
had something like 15-20 026s and a 407 setup to print card decks. At
the end of the semester was hired to rewrite 1401 MPIO in 360
assembler for 360/30. The univ was getting 360/67 replacing 709/1401
for tss/360, 360/30 temporarily replaced 1401 pending arrival of
360/67. Univ. shutdown datacenter on weekends and I had whole place
dedicated (although 48hrs w/o sleep made monday classes difficult). I
was given pile of hardware and software manuals and got to design and
implement my own monitor, device drivers, interrupt handlers, error
recovery, storage management, etc. and after a few weeks had nearly
full tray of (2000) cards. Sometimes come in and production had
finished early and place was dark with everything shut off and
couldn't power up 360/30. Trial and error learned to put the
controllers in CE-mode, power on 360/30, power on controllers and then
take controllers out of CE-mode.
The 360/67 arrived within a year of taking intro class and I was hired
fulltime responsible for os/360 (tss/360 not coming to
production). 709 did student fortran in less than second. 360/67 (as
360/65) os/360 (MFT9.5) took over a minute, I install HASP cutting
time in half. MFT11, I start redoing SYSGEN STAGE2 (partial tray) to
carefully place datasets and PDS members to optimize arm seek and (PDS
directory) multi-track search, cutting another 2/3rds to
12.9secs. OS/360 never got better than 709 until I install Univ
Waterloo WATFOR; as 360/65, WATFOR clocked at 20,000 cards/min
(333cards/sec), student fortran tended to run 30-60 statements.
Then IBM CSC came out to install (virtual machine, precursor to vm370)
CP/67 (3rd after CSC itself and MIT Lincoln Labs) and I mostly get to
play with it during my weekend 48hr window. I then spend a few months
rewriting pathlengths for running OS/360 in virtual machine. Source
was in OS/360 files which had to be modified, assembled, mark txt deck
and merge it into a tray of txt decks (that was CP67 executable). BPS
loader IPL the tray of TXT decks which then writes core image to disk,
for ipl'ing the system.
Bare machine test ran 322secs ... initially 856secs (CP67 CPU
534secs). After a few months I had CP67 CPU down from 534secs to
113secs. I then start rewriting the dispatcher/scheduler , (dynamic
adaptive resource manager/default fair share scheduling policy),
paging, adding ordered seek queuing (from FIFO) and mutli-page
transfer channel programs (from FIFO and optimized for
transfers/revolution, getting 2301 paging drum from 70-80 4k
transfers/sec to channel transfer peak of 270).
Six months after univ initial CP/67 install, CSC was giving one week
class in LA. I arrive on Sunday afternoon and asked to teach the
class, it turns out that the people that were going to teach it had
resigned the Friday before to join one of the 60s CSC CP67 online
commercial spin-offs.
Before I graduate, I'm hired fulltime into small group in Boeing CFO
office to help with formation of Boeing Computer Services (consolidate
all dataprocessing into independent business unit). I think Renton,
largest datacenter in the world, 360/65s arriving faster than they
could be installed, boxes constantly staged in hallways around machine
room (did have lone 360/75, when doing classified work; black rope
around area, guards at corners, heavy black felt covering console
lights and visible 1403 printer areas).
Lots of politics between Renton director and CFO, who only had a
360/30 up at Boeing Field for payroll, although they enlarge the room
to install 360/67 for me to play with when I wasn't doing other
stuff. I had mid-level management badge so could park in lots closer
to the bldgs. 747-3 was flying skies of Seattle getting FAA flt
certification. Tours of mock-up of 747 cabin just south of Boeing
field would claim 747s carried so many passengers, they would never be
serviced with fewer than four jetways.
Both Boeing and IBM told stories about 360 announcement day, Boeing
submitted an order making the IBM marketing rep, highest paid IBM
employee that year. Sales were still straight commission; IBM shifts
to quota the following year.
some recent posts mentionin 709 1401 mpio fortran watfor cp/67
csc boeing cfo
https://www.garlic.com/~lynn/2026c.html#22 NSFNET
https://www.garlic.com/~lynn/2026b.html#78 Early CSC (& CP67/CMS)
https://www.garlic.com/~lynn/2026b.html#73 IBM System/360
https://www.garlic.com/~lynn/2026b.html#70 IBM Online Systems
https://www.garlic.com/~lynn/2026b.html#14 Bad Responsee
https://www.garlic.com/~lynn/2026b.html#4 IBM Virtual Machine
https://www.garlic.com/~lynn/2026.html#98 IBM 360&370 Experience
https://www.garlic.com/~lynn/2026.html#91 CP/67 and VM/370
https://www.garlic.com/~lynn/2026.html#83 Touch Type, Typewriters, Terminals
https://www.garlic.com/~lynn/2026.html#82 IBM DASD, CKD, FBA
https://www.garlic.com/~lynn/2026.html#78 IBM OS Debugging
https://www.garlic.com/~lynn/2026.html#67 Early Mainframe work
https://www.garlic.com/~lynn/2026.html#59 IBM CP67 and VM370
https://www.garlic.com/~lynn/2026.html#28 360 Channel
https://www.garlic.com/~lynn/2026.html#24 IBM 360, Future System
https://www.garlic.com/~lynn/2025e.html#104 Early Mainframe Work
https://www.garlic.com/~lynn/2025e.html#74 IBM 370 Virtual Memory
https://www.garlic.com/~lynn/2025e.html#57 IBM 360/30 and other 360s
https://www.garlic.com/~lynn/2025d.html#112 Mainframe and Cloud
https://www.garlic.com/~lynn/2025d.html#99 IBM Fortran
https://www.garlic.com/~lynn/2025d.html#15 MVT/HASP
https://www.garlic.com/~lynn/2025c.html#115 IBM VNET/RSCS
https://www.garlic.com/~lynn/2025c.html#64 IBM Vintage Mainframe
https://www.garlic.com/~lynn/2025c.html#55 Univ, 360/67, OS/360, Boeing, Boyd
https://www.garlic.com/~lynn/2025.html#91 IBM Computers
https://www.garlic.com/~lynn/2024g.html#17 60s Computers
https://www.garlic.com/~lynn/2024f.html#110 360/65 and 360/67
https://www.garlic.com/~lynn/2024f.html#88 SHARE User Group Meeting October 1968 Film Restoration, IBM 360
https://www.garlic.com/~lynn/2024f.html#69 The joy of FORTH (not)
https://www.garlic.com/~lynn/2024f.html#20 IBM 360/30, 360/65, 360/67 Work
https://www.garlic.com/~lynn/2024d.html#103 IBM 360/40, 360/50, 360/65, 360/67, 360/75
https://www.garlic.com/~lynn/2024d.html#76 Some work before IBM
https://www.garlic.com/~lynn/2024c.html#15 360&370 Unix (and other history)
https://www.garlic.com/~lynn/2024b.html#97 IBM 360 Announce 7Apr1964
https://www.garlic.com/~lynn/2024b.html#63 Computers and Boyd
https://www.garlic.com/~lynn/2024b.html#60 Vintage Selectric
https://www.garlic.com/~lynn/2024b.html#44 Mainframe Career
https://www.garlic.com/~lynn/2023e.html#64 Computing Career
https://www.garlic.com/~lynn/2023e.html#34 IBM 360/67
https://www.garlic.com/~lynn/2023d.html#106 DASD, Channel and I/O long winded trivia
https://www.garlic.com/~lynn/2023d.html#88 545tech sq, 3rd, 4th, & 5th flrs
https://www.garlic.com/~lynn/2023d.html#32 IBM 370/195
https://www.garlic.com/~lynn/2023c.html#82 Dataprocessing Career
https://www.garlic.com/~lynn/2023c.html#73 Dataprocessing 48hr shift
https://www.garlic.com/~lynn/2022h.html#31 IBM OS/360
https://www.garlic.com/~lynn/2022e.html#95 Enhanced Production Operating Systems II
https://www.garlic.com/~lynn/2022.html#12 Programming Skills
--
virtualization experience starting Jan1968, online at home since Mar1970
MIT CTSS/7094, MULTICS, UNIX, Virtual Machines
From: Lynn Wheeler <lynn@garlic.com>
Subject: MIT CTSS/7094, MULTICS, UNIX, Virtual Machines
Date: 02 Sep, 2026
Blog: Facebook
Some of the MIT CTSS/7094 went to the 5th flr to do MULTICS.
https://en.wikipedia.org/wiki/Multics
Due to its many novel and valuable ideas, Multics has had a
significant influence on computer science despite its faults.[7] Its
most lasting effect on the computer industry was to inspire the
creation of Unix, which carried forward many Multics features, but was
able to run on less expensive hardware. Unix was developed at Bell to
allow their Multics team to continue their research using smaller
machines, first a PDP-7 and ultimately the PDP-11.
... snip ...
Others went to the IBM Cambridge Scientific Center on the 4th flr and
did virtual machines (wanted 360/50 to modify with virtual memory but
all the extra were going to FAA ATC, so had to settle for 360/40 to
modify and did CP40/CMS, which morphs into CP67/CMS when 360/67
standard with virtual memory became available, precursor to
VM370/CMS).
CTSS RUNOFF was rewritten for CMS as SCRIPT. Then GML was invented at
CSC in 1969 and GML tag processing was added to SCRIPT (decade later
morphs into SGML, and after another decade morphs into HTML at CERN)
MIT CTSS /7094 TYPSET, RUNOFF
https://en.wikipedia.org/wiki/TYPSET_and_RUNOFF
The origin of IBM's SCRIPT software began in 1968 when IBM contracted
Stuart Madnick of MIT to write a simple document preparation tool[7]
for CP/67,[8] which he modelled on MIT's CTSS RUNOFF.[9]
... snip ...
CTSS/7094 email
https://www.multicians.org/thvv/anhc-34-1-anec.html
History of Electronic Mail
https://www.multicians.org/thvv/mail-history.html
IBM CP/CMS had electronic mail as early as 1966, and was widely used
within IBM in the 1970s. Eventually this facility evolved into the
PROFS product (more like spinoff) in the 1980s.
... snip ...
CSC co-worker responsible for the CP67-based wide-area network (morphs
into the internal corporate network, larger than arpanet/internet from
the beginning until sometime mid/late 80s about time forced to convert
to SNA/VTAM) ... comment by one of the 1969 GML inventors:
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for
the project, something I was allowed to do part-time because of my
knowledge of the user requirements. My real job was to encourage the
staffs of the various scientific centers to make use of the
CP-67-based Wide Area Network that was centered in Cambridge.
... snip ...
Edson was responsible for CP67 wide-area network
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which
paved the way to the Internet as we know it today.
... snip ...
Technology also used for the corporate sponsored Univ. BITNET
https://en.wikipedia.org/wiki/BITNET
IBM 4300 sold in same mid-range market as VAX and in about the same
numbers for single or small unit orders. Big difference starting in
early 80s was large corporations ordering hundreds of 4300s at a time
for distributing out in departmental areas (inside IBM conference
rooms became scarce so many converted to VM/4300 rooms). Old archive
post with a decade of VAX orders sliced and diced by model, year,
US/non-US:
https://web.archive.org/web/20260305143949/https://www.garlic.com/~lynn/2002f.html#0
Late 70s, I transferred out to SJR on the west coast, worked with Jim
Gray and Vera Watson on original SQL/Relational, System/R and also got
to wander around datacenters in silicon valley, including disk
bldg14/engineering and bldg15/product test across the street. At the
time, they were running 7x24, prescheduled, stand-alone testing and
mentioned that they had recently tried "MVS", but it had 15min MTBF
(in that environment). I offer to rewrite the I/O supervisor to make
it bullet proof and never fail, allowing any amount of concurrent,
on-demand testing, greatly improving productivity. Bldg15 then gets
1st engineering 3033 (outside POK processor engineering). Then summer
of 1978, bldg15 gets engineering 4341 (year before customer
ship). Jan1979, branch office hears about it and cons me into doing a
(CDC6600) Fortran benchmark for national lab looking at getting 70 for
compute farm (sort of the leading edge of the coming cluster
supercomputing tsunami).
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
GML, SGML, HTML posts
https://www.garlic.com/~lynn/submain.html#sgml
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
bitnet posts
https://www.garlic.com/~lynn/subnetwork.html#bitnet
original SQL/relational, System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
getting to play disk engineer posts
https://www.garlic.com/~lynn/subtopic.html#disk
--
virtualization experience starting Jan1968, online at home since Mar1970
Internet Lore
From: Lynn Wheeler <lynn@garlic.com>
Subject: Internet Lore
Date: 02 Sep, 2026
Blog: Facebook
Early 80s, I got "HSDT", T1 and faster computer links (terrestrial and
satellite) and battles with the communication group (60s had 2701 that
supported T1 links, 70s issues with VTAM capped controllers at
56kbits). Funding was predicated on being able to show some IBM
content, was able to find FSD S/1 T1 Zirpel cards (for gov. agency
2701s that were failing).
Also working with NSF director and was suppose to get $20M to
interconnect the NSF supercomputer centers. NSF gives grant to UC for
Berkeley Supercomputer Center and I'm asked to give them presentations
on the interconnect. However, UC Regents plan for next new bldg was
for UCSD and it becomes the San Diego Supercomputer Center instead,
run by General Atomics (NSF didn't have control). Then congress cuts
the budget, some other things happened and eventually there was RFP
released (in part based on what we already had running). NSF 28Mar1986
Preliminary Announcement:
https://web.archive.org/web/20260311074443/https://www.garlic.com/~lynn/2002k.html#12
The OASC has initiated three programs: The Supercomputer Centers
Program to provide Supercomputer cycles; the New Technologies Program
to foster new supercomputer software and hardware developments; and
the Networking Program to build a National Supercomputer Access
Network - NSFnet.
...
IBM internal politics was not allowing us to bid. The NSF director
tried to help by writing the company a letter (3Apr1986, NSF Director
to IBM Chief Scientist and IBM Senior VP and director of Research,
copying IBM CEO) with support from other gov. agencies ... but that
just made the internal politics worse (as did claims that what we
already had operational was at least 5yrs ahead of the winning bid),
as regional networks connect in, NSFnet becomes the NSFNET backbone,
precursor to modern internet.
1988, Nick Donofrio approves HA/6000, initially for NYTimes to move
their newspaper system (ATEX) off DEC VAXCluster to RS/6000, I rename
it HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing (technical/scientific) cluster scale-up with
national labs (LANL, LLNL, NCAR/UCAR, etc) and commercial cluster
scale-up with RDBMS vendors (Oracle, Sybase, Ingres, Informix) with
VAXCluster support in same source base with UNIX. Also migrate LLNL
UNICOS "LINCS"
https://pdfs.semanticscholar.org/f5fc/36f760a62a39e1b780b556d5817841d40fac.pdf
and NCAR "Mesa Archive" spin-off, to HA/CMP.
Early Jan92, there was meeting with Oracle CEO and IBM/AWD executive
Hester tells Ellison that we would have 16-system clusters by mid92
and 128-system clusters by ye92. Mid-jan92, I update FSD on HA/CMP
work with national labs and FSD decides to go with HA/CMP for federal
supercomputers. By end of Jan, we are told that cluster scale-up is
being transferred to Kingston for announce as IBM Supercomputer (SP,
technical/scientific *ONLY*) and we aren't allowed to work with
anything that has more than four systems (we leave IBM a few months
later). A couple weeks later, 17feb1992, Computerworld news ... IBM
establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
After leaving IBM was brought into small client/server startup. Two
former Oracle people (that had been in the Ellison/Hester meeting)
were there responsible for something called "commerce server" and they
wanted to be able to do payment transactions. The startup had also
invented some technology called SSL they wanted to use, the result is
now sometimes called e-commerce. I was responsible for everything
between webservers and payment networks (trivia: payment network
trouble desks had requirement that 1st level problem determination be
done within 5mins, webservers was more like 3hrs and NTF). I then did
a talk on "Why Internet Wasn't Business Critical Dataprocessing"
(based on documentation, software, processes I had do for financial
transactions); IETF RFC Editor Postel sponsored my talk at USC/ISI.
One of the CSC members was responsible for Science Center CP67-based
wide-area network; later morphs into the IBM internal network (larger
than arpanet/internet from beginning until sometime mid/late 80s when
it was forced to convert to SNA/VTAM) and technology also used for the
corporate sponsored UNIV "BITNET".
https://en.wikipedia.org/wiki/BITNET
From one of the CSC co-inventers of GML (precursor to SGML & HTML) in
1969:
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for
the project, something I was allowed to do part-time because of my
knowledge of the user requirements. My real job was to encourage the
staffs of the various scientific centers to make use of the
CP-67-based Wide Area Network that was centered in Cambridge.
...
Edson
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which
paved the way to the Internet as we know it today.
...
SJMerc article about Edson (he passed aug2020) and "IBM'S MISSED
OPPORTUNITY WITH THE INTERNET" (gone behind paywall but lives free at
wayback machine)
https://web.archive.org/web/20000124004147/http://www1.sjmercury.com/svtech/columns/gillmor/docs/dg092499.htm
Also from wayback machine, some additional (IBM missed internet)
references from Ed's website
https://web.archive.org/web/20000115185349/http://www.edh.net/bungle.htm
Edson and I transfer out to SJR on the west coast in 1977.
Growth of ARPANET appeared to be inhibited by requiring IMPs (at
Jan1983 cutover something like 100 IMPs and 255 hosts). Growth of
corporate internal network was inhibited by requirement that all links
have encryptors and gov. resistance, especially when crossed national
boundaries (1983, passed 1000 nodes). Archive post with world-wide
corporate locations that added one or more nodes during 1983:
https://web.archive.org/web/20251204205713/https://www.garlic.com/~lynn/2006k.html#8
First webserver in US was on the SLAC (IBM) VM370
https://ahro.slac.stanford.edu/wwwslac-exhibit
https://ahro.slac.stanford.edu/wwwslac-exhibit/early-web-chronology-and-documents-1991-1994
SJR was 1st to put up (non-IBM) gateway to CSNET in Oct1982:
Date: 30 Dec 1982 14:45:34 EST (Thursday)
From: Nancy Mimno <mimno@Bbn-Unix>
Subject: Notice of TCP/IP Transition on ARPANET
To: csnet-liaisons at Udel-Relay
Cc: mimno at Bbn-Unix
Via: Bbn-Unix; 30 Dec 82 16:07-EST
Via: Udel-Relay; 30 Dec 82 13:15-PDT
Via: Rand-Relay; 30 Dec 82 16:30-EST
ARPANET Transition 1 January 1983
Possible Service Disruption
Dear Liaison,
As many of you may be aware, the ARPANET has been going through the
major transition of shifting the host-host level protocol from NCP
(Network Control Protocol/Program) to TCP-IP (Transmission Control
Protocol - Internet Protocol). These two host-host level protocols are
completely different and are incompatible. This transition has been
planned and carried out over the past several years, proceeding from
initial test implementations through parallel operation over the last
year, and culminating in a cutover to TCP-IP only 1 January 1983. DCA
and DARPA have provided substantial support for TCP-IP development
throughout this period and are committed to the cutover date.
The CSNET team has been doing all it can to facilitate its part in
this transition. The change to TCP-IP is complete for all the CSNET
host facilities that use the ARPANET: the CSNET relays at Delaware and
Rand, the CSNET Service Host and Name Server at Wisconsin, the CSNET
CIC at BBN, and the X.25 development system at Purdue. Some of these
systems have been using TCP-IP for quite a while, and therefore we
expect few problems. (Please note that we say "few", not "NO
problems"!) Mail between Phonenet sites should not be affected by the
ARPANET transition. However, mail between Phonenet sites and ARPANET
sites (other than the CSNET facilities noted above) may be disrupted.
The transition requires a major change in each of the more than 250
hosts on the ARPANET; as might be expected, not all hosts will be
ready on 1 January 1983. For CSNET, this means that disruption of mail
communication will likely result between Phonenet users and some
ARPANET users. Mail to/from some ARPANET hosts may be delayed; some
host mail service may be unreliable; some hosts may be completely
unreachable. Furthermore, for some ARPANET hosts this disruption may
last a long time, until their TCP-IP implementations are up and
working smoothly. While we cannot control the actions of ARPANET
hosts, please let us know if we can assist with problems, particularly
by clearing up any confusion. As always, we are <cic@csnet-sh> or
(617)497-2777.
Please pass this information on to your users.
Respectfully yours,
Nancy Mimno
CSNET CIC Liaison
---
Date: 02/02/83 23:49:45
To: CSNET mailing list
Subject: CSNET headers, CSNET status
You may have noticed that since ARPANET switched to TCP/IP and the new
version of software on top of it, message headers have become
ridiculously long. Some of it is because of tracing information that
has been added to facilitate error isolation and "authentication", and
some of it I think is a bug (the relay adds a 'From' and a 'Date'
header although there already are headers with that information in the
message). This usually doesn't bother people on the ARPANET because
they have smart mail reading programs that understand the headers and
only display the relevant ones. I have proposed a mail reader/sender
program that understands about ARPANET headers (RFC822) as a summer
project, so maybe we will sometime enjoy the same priviledge.
The file CSNET STATUS1 on the CSNET disk (see instructions below for
how to access it) contains some clarification of the problems that
have been experienced with the TCP/IP conversion. Here is a summary:
- Nodes that don't yet talk TCP (but the old NCP) can be accessed
through the UDel-Relay. So if you think you have problems reaching
a node because of this, append @Udel-Relay to the ARPANET
- You can find out about the status of hosts (e.g., if they run TCP or
not) by sending ANY MESSAGE to Status@UDel-Relay (capitalization is
NOT significant).
- If your messages are undeliverable, you get a notice after two days,
and your messages get returned after 4 days.
- Avoid using any of the fancy address forms allowed by the new header
format (RFC822).
- The TCP transition was a lot more trouble than the ARPANET people
had anticipated.
HSDT posts
https://www.garlic.com/~lynn/subnetwork.html#hsdt
NSFNET posts
https://www.garlic.com/~lynn/subnetwork.html#nsfnet
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
Ecommerce gateway posts
https://www.garlic.com/~lynn/subnetwork.html#gateway
BITNET posts
https://www.garlic.com/~lynn/subnetwork.html#bitnet
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
GML, SGML, HTML posts
https://www.garlic.com/~lynn/submain.html#sgml
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
--
virtualization experience starting Jan1968, online at home since Mar1970
Cloud Computing
From: Lynn Wheeler <lynn@garlic.com>
Subject: Cloud Computing
Date: 03 Sep, 2026
Blog: Facebook
Some of the MIT CTSS/7094
https://en.wikipedia.org/wiki/Compatible_Time-Sharing_System
people went to the 5th flr and MULTICS
https://en.wikipedia.org/wiki/Multics
others went to the 4th flr and IBM Cambridge Science Center
https://en.wikipedia.org/wiki/Cambridge_Scientific_Center
CSC wanted 360/50 to modify with virtual memory, but all the spare
360/50s were going to FAA ATC, so had to settle for 360/40 to modify
and did (virtual machine) CP40/CMS, which morphs into CP67/CMS when
360/67 standard with virtual memory becomes available.
As undergraduate at univ, I was hired fulltime responsible for OS/360
for 360/67 (running as 360/65, tss/360 not coming to production)
Then IBM CSC came out to install (virtual machine) CP/67 (3rd
after CSC itself and MIT Lincoln Labs) and I mostly get to play with
it during my weekend 48hr window. I then spend a few months rewriting
pathlengths for running OS/360 in virtual machine. Bare machine
test ran 322secs ... initially 856secs (CP67 CPU 534secs). After a few
months I had CP67 CPU down from 534secs to 113secs. I then start
rewriting the dispatcher/scheduler, (dynamic adaptive resource
manager/default fair share scheduling policy), paging, adding ordered
seek queuing (from FIFO) and mutli-page transfer channel programs
(from FIFO and optimized for transfers/revolution, getting 2301 paging
drum from 70-80 4k transfers/sec to channel transfer peak of 270).
Six months after univ initial CP/67 install, CSC was giving one week
class in LA. I arrive on Sunday afternoon and asked to teach the
class, it turns out that the people that were going to teach it had
resigned the Friday before, to join one of the 60s CSC CP67 commercial
online spin-offs.
CSC and CSC commercial online spin-offs do a lot of CP67 automation
work in the 60s, being able to operate offshift and weekends
unattended.
Before graduation, I was hired into small group in Boeing CFO office
to help with the formation Boeing Computer Services (consolidate all
dataprocessing into independent business unit). I think Renton
datacenter largest in the world. Both Boeing and IBMers tell story
about the day IBM System/360 was announced (7Apr1964), Boeing walks in
with an order making the marketing rep, highest paid IBM employee that
year (days of straight commission, following year, IBM changes to
"quota").
Current cloud operations can have a dozen or more megadatacenters
around world, each with half million or more server blades with
enormous automation. Each megadatacenter operating with 70-80 staff.
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
cloud megadatacenter posts
https://www.garlic.com/~lynn/submisc.html#megadatacenter
--
virtualization experience starting Jan1968, online at home since Mar1970
Cloud Computing
From: Lynn Wheeler <lynn@garlic.com>
Subject: Cloud Computing
Date: 04 Sep, 2026
Blog: Facebook
re:
https://www.garlic.com/~lynn/2026c.html#34 Cloud Computing
When I graduated, I join CSC (instead of staying with Boeing CFO) and
one of my hobbies was enhanced production operating systems for
internal datacenters. One of the first (and long time) internal
customers was the online branch office sales&marketing support HONE
systems; initially several CP67 datacenters around the US. I move to
enhanced VM370 and HONE is migrated to VM370 and all the US HONE
datacenters are consolidated in Palo Alto and all the VM370 systems
are tied together into "single-system image", loosely-coupled, shared
DASD complex with load balancing across the whole operation (when
FACEBOOK 1st moves into silicon valley it is into a new bldg built
next door to the former consolidated US HONE datacenter) and country
HONE systems were sprouting up all over the world.
Large cloud operations providing online, on-demand services will have
frequently significantly provisioned for peak demand ten times avg
which can result it frequent, large idle resources. Early last decade
it was possible to use credit card at cloud megadatacenters to spin-up
large supercomputer systems (that would rank in top 40 in the world)
during low cloud demand.
High overlap between cloud megadatacenters and supercomputer
technologies.
Trivia: Second half 70s, I transferred out to SJR on the west coast,
worked with Jim Gray and Vera Watson on original SQL/Relational,
System/R and also got to wander around datacenters in silicon valley,
including disk bldg14/engineering and bldg15/product test across the
street. At the time, they were running 7x24, prescheduled, stand-alone
testing and mentioned that they had recently tried "MVS", but it had
15min MTBF (in that environment). I offer to rewrite the I/O
supervisor to make it bullet proof and never fail, allowing any amount
of concurrent, on-demand testing, greatly improving productivity.
Bldg15 then gets 1st engineering 3033 (outside POK processor
engineering). Then summer of 1978, bldg15 gets engineering 4341 (year
before customer ship). Jan1979, branch office hears about it and cons
me into doing a (CDC6600) Fortran benchmark for national lab looking
at getting 70 VM/4341s for compute farm (sort of the leading edge of
the coming cluster supercomputing tsunami).
Decade later, 1988 HA/6000 is approved, originally for NYTimes to move
their newspaper system ("ATEX") off DEC VAXCluster to RS/6000 (running
project out at IBM Los Gatos lab). I rename it HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing technical/scientific cluster scale-up with national
labs (LANL, LLNL, NCAR, etc) and commercial cluster scale-up with
RDBMS vendors (Oracle, Sybase, Ingres, Informix) with VAXCluster
support in same source base with Unix. I was planning on using FCS for
HA/CMP and ported LLNL's UNICOS LINCS and NCAR's filesystem (spun off
as Mesa Archival) to HA/CMP.
Early Jan92, there was HA/CMP meeting with Oracle CEO and IBM/AWD
executive Hester tells Ellison that we would have 16-system clusters
by mid92 and 128-system clusters by ye92. Mid-jan92, I update FSD on
HA/CMP work with national labs and FSD decides to go with HA/CMP for
federal supercomputers. By end of Jan92, we are told that cluster
scale-up is being transferred to Kingston for announce as IBM
Supercomputer (SP, technical/scientific *ONLY*) and we aren't allowed
to work with anything that has more than four systems (we leave IBM a
few months later). A couple weeks later, 17feb1992, Computerworld news
... IBM establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
HONE posts
https://www.garlic.com/~lynn/subtopic.html#hone
original SQL/relational RDBMS, System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
cloud megadatacenter posts
https://www.garlic.com/~lynn/submisc.html#megadatacenter
Getting to get to play disk engineer posts
https://www.garlic.com/~lynn/subtopic.html#disk
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
--
virtualization experience starting Jan1968, online at home since Mar1970
VM/370
From: Lynn Wheeler <lynn@garlic.com>
Subject: VM/370
Date: 06 Sep, 2026
Blog: Facebook
In 1974, CERN had presented a comparison of MVS/TSO with VM370/CMS at
SHARE. The presentation was freely available outside IBM, but inside
IBM, copies were stamped "IBM Confidential - Restricted" (available on
need to know only). Some speculation it contributed to head of IBM POK
convincing corporate to kill VM370 product, shutdown the development
group and transfer all the people to POK to work on MVS/XA.
They were planning on not telling the people until the very last
moment (to minimize those that might escape into the Boston area). It
managed to leak early and several managed to escape (there was hunt to
find leak source, fortunately for me, nobody gave them up). This was
in DEC VAX/VMS infancy and joke was that head of IBM POK was major
contributor to VMS. As mentioned several times, Endicott managed to
save the VM370 product mission, but had to recreate a development
group from scratch.
Co-worker at CSC was responsible for the CP67-based scientific center
"wide-area" network (morphs into RSCS/VNET and the IBM internal
network, larger than ARPANET/Internet from the beginning until
sometime mid/late 80s; about the same time as internal network was
forced to convert to SNA/VTAM). Technology also used for the corporate
sponsored univ. BITNET.
Account by one of the CSC inventors of GML in 1969 (a decade later,
GML morphs into SGML, and after another decade morphs into HTML at
CERN):
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for
the project, something I was allowed to do part-time because of my
knowledge of the user requirements. My real job was to encourage the
staffs of the various scientific centers to make use of the
CP-67-based Wide Area Network that was centered in Cambridge.
... snip ...
Edson was responsible for CP67 wide-area network
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which
paved the way to the Internet as we know it today.
... snip ....
SJMerc article about Edson (he passed aug2020) and "IBM'S MISSED
OPPORTUNITY WITH THE INTERNET" (gone behind paywall but lives free at
wayback machine)
https://web.archive.org/web/20000124004147/http://www1.sjmercury.com/svtech/columns/gillmor/docs/dg092499.htm
Also from wayback machine, some additional (IBM missed internet)
references from Ed's website
https://web.archive.org/web/20000115185349/http://www.edh.net/bungle.htm
To allow MVS into the corporate network, a HASP/JES2 emulated driver
was done for RSCS/VNET. However HASP/JES2 was very fragile. The
original HASP network code used to have "TUCC" in cols78-81. It used
spare entries in the 255 entry psuedo device table (usually around
160-180 and HASP/JES2 would trash traffic that didn't have origin or
destination entries in local table) however the internal network had
long based 255 nodes.
Later JES2 had habit of crashing MVS when processing traffic at
different JES2 release level. Lots of code was added to the RSCS/VNET
JES2 emulated driver to recognize traffic originating from JES2 at
different release level from the directly connected JES2 destination
(and rewrite the header info). In the 80s, JES2 was enhanced to handle
999 nodes, but the internal network had already passed 1000.
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
internetal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
BITNET posts
https://www.garlic.com/~lynn/subnetwork.html#bitnet
posts mention CERN MVS/TS0-VM370/CMS analysis
https://www.garlic.com/~lynn/2026b.html#9 CMS, Self-hosting and the 6502
https://www.garlic.com/~lynn/2026b.html#4 IBM Virtual Machine
https://www.garlic.com/~lynn/2026b.html#2 IBM 370/195
https://www.garlic.com/~lynn/2026.html#47 IBM Internal Network
https://www.garlic.com/~lynn/2026.html#38 IBM CMS Applications
https://www.garlic.com/~lynn/2025e.html#103 IBM CSC, HONE
https://www.garlic.com/~lynn/2025e.html#6 VM370 Teddy Bear
https://www.garlic.com/~lynn/2025d.html#104 Rapid Response
https://www.garlic.com/~lynn/2025d.html#59 IBM Example Programs
https://www.garlic.com/~lynn/2025d.html#37 TYMSHARE, VMSHARE, ADVENTURE
https://www.garlic.com/~lynn/2025d.html#34 IBM Internal Apps, Retain, HONE, CCDN, ITPS, Network
https://www.garlic.com/~lynn/2025d.html#18 Some VM370 History
https://www.garlic.com/~lynn/2025d.html#6 SLAC and CERN
https://www.garlic.com/~lynn/2025c.html#0 Interactive Response
https://www.garlic.com/~lynn/2024.html#90 IBM, Unix, editors
https://www.garlic.com/~lynn/2023d.html#16 Grace Hopper (& Ann Hardy)
https://www.garlic.com/~lynn/2022h.html#69 Fred P. Brooks, 1931-2022
https://www.garlic.com/~lynn/2022g.html#56 Stanford SLAC (and BAYBUNCH)
https://www.garlic.com/~lynn/2022d.html#60 VM/370 Turns 50 2Aug2022
https://www.garlic.com/~lynn/2010q.html#34 VMSHARE Archives
--
virtualization experience starting Jan1968, online at home since Mar1970
Lynn Conway
From: Lynn Wheeler <lynn@garlic.com>
Subject: Lynn Conway
Date: 09 Sep, 2026
Blog: Facebook
wasn't aware, until other prior references here on facebook
Lynn Conway, 1938-2024: The Computer Architect Who Helped to
Revolutionize Digital IC Design
https://www.eejournal.com/article/lynn-conway-1938-2024-the-computer-architect-who-helped-to-revolutionize-digital-ic-design/
Lynn Conway
https://en.wikipedia.org/wiki/Lynn_Conway
In Memoriam: Lynn Conway (1938-2024)
https://computerhistory.org/blog/in-memoriam-lynn-conway-1938-2024/
... I had transferred out to San Jose Research in 1977 and in early
80s, the Los Gatos lab also let me have part of a wing out there with
the VLSI group (and I never encountered any reference) ... but ACS had
been killed in 1969 (before I graduated and joined IBM, folklore was
executives felt it would advance the state-of-art too fast and loose
control of the market) and Amdahl left IBM:
https://people.computing.clemson.edu/~mark/acs_end.html
trivia: shortly after joining IBM at the Cambridge Scientific Center,
the 370/195 group cons me into helping with hyperthreading the
machine. 195 had pipelining and out of order execution, but
conditional branches drained the pipeline and most codes only ran at
half rated speed. Two instruction streams (??) could keep execution
units busy. Then it was canceled when decision was made to add virtual
memory to all 370s, and it was decided it would be too hard to
retrofit virtual memory to 195.
Hyperthreading 195 wouldn't have worked very well, it needed MVT 2-CPU
multiprocessor support which had very high overhead and wouldn't
achieve anything close to two times throughput.
... then there is this (having dealt a lot with TYMSHARE): Ann Hardy
https://medium.com/chmcore/someone-elses-computer-the-prehistory-of-cloud-computing-bca25645f89
Ann Hardy is a crucial figure in the story of Tymshare and
time-sharing. She began programming in the 1950s, developing software
for the IBM Stretch supercomputer. Frustrated at the lack of
opportunity and pay inequality for women at IBM -- at one point she
discovered she was paid less than half of what the lowest-paid man
reporting to her was paid -- Hardy left to study at the University of
California, Berkeley, and then joined the Lawrence Livermore National
Laboratory in 1962. At the lab, one of her projects involved an early
and surprisingly successful time-sharing operating system.
Ann Hardy at Computer History Museum
https://www.computerhistory.org/collections/catalog/102717167
Ann rose up to become Vice President of the Integrated Systems
Division at Tymshare, from 1976 to 1984, which did online airline
reservations, home banking, and other applications. When Tymshare was
acquired by McDonnell-Douglas in 1984, Ann's position as a female VP
became untenable, and was eased out of the company by being encouraged
to spin out Gnosis, a secure, capabilities-based operating system
developed at Tymshare. Ann founded Key Logic, with funding from Gene
Amdahl, which produced KeyKOS, based on Gnosis, for IBM and Amdahl
mainframes. After closing Key Logic, Ann became a consultant, leading
to her cofounding Agorics with members of Ted Nelson's Xanadu project
... snip ...
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
SMP, tightly-coupled, shared memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
past posts mentioning Lynn Conway
https://www.garlic.com/~lynn/2021d.html#35 April 7, 1964: IBM Bets Big on System/360
https://www.garlic.com/~lynn/2007t.html#57 Unsung innovators: 10 people who shaped the computer industry
https://www.garlic.com/~lynn/2003j.html#18 why doesn't processor reordering instructions affect most
a few past posts mentioning Ann Hardy
https://www.garlic.com/~lynn/2025e.html#56 Tymshare
https://www.garlic.com/~lynn/2024f.html#75 Prodigy
https://www.garlic.com/~lynn/2024f.html#11 TYMSHARE, Engelbart, Ann Hardy
https://www.garlic.com/~lynn/2024e.html#1 TYMSHARE Dialup
https://www.garlic.com/~lynn/2024d.html#78 Other Silicon Valley
https://www.garlic.com/~lynn/2024d.html#47 E-commerce
https://www.garlic.com/~lynn/2024c.html#25 Tymshare & Ann Hardy
https://www.garlic.com/~lynn/2023e.html#62 IBM Jargon
https://www.garlic.com/~lynn/2023e.html#12 Tymshare
https://www.garlic.com/~lynn/2023e.html#9 Tymshare
https://www.garlic.com/~lynn/2023d.html#37 Online Forums and Information
https://www.garlic.com/~lynn/2023d.html#16 Grace Hopper (& Ann Hardy)
https://www.garlic.com/~lynn/2023c.html#97 Fortran
https://www.garlic.com/~lynn/2023b.html#35 When Computer Coding Was a 'Woman's' Job
--
virtualization experience starting Jan1968, online at home since Mar1970
Reliability, Availability, Serviceability
From: Lynn Wheeler <lynn@garlic.com>
Subject: Reliability, Availability, Serviceability
Date: 11 Sep, 2026
Blog: Facebook
I got coned into helping with a 16-CPU 370 multiprocessor effort and
we talk the 3033 processor engineers into working on it in their spare
time (a lot more interesting than remapping 168 logic to 20% faster
chips). Everybody thought it was great until somebody tells the head
of POK that it could be decades before POK's favorite son operating
system ("MVS") has (effective) 16-CPU support (POK doesn't ship 16-CPU
system until after turn of century; IBM documents at the time was that
MVS 2-CPU support only got 1.2-1.5 throughput of single processor
operation). Then some of us were invited to never visit POK again and
the 3033 processor engineers were directed, heads down and no
distractions (once 3033 is out the door, the 3033 processor engineers
start on trout/3090).
After being invited to never visit POK again, I transfer out to SJR on
the west coast and worked with Jim Gray and Vera Watson on the
original SQL/Relational implementation, System/R (some resistance from
the STL next great DBMS "EAGLE"). Was able to do tech transfer to
Endicott ("under the radar" while company was preoccupied by "EAGLE")
for SQL/DS. Then when "EAGLE" imploded was asked how fast could it be
ported to MVS, eventually released as DB2, originally for "decision
support" *ONLY*.
I also got to wander around silicon valley datacenters, including disk
bldg14/engineering and bldg15/product test, across the street. At the
time they were doing pre-scheduled, around the clock stand-alone
testing. They mentioned that recently had tried MVS, but it had 15min
MTBF (in that environment). I offered to rewrite I/O Supervisor to
make it bullet-proof and never fail, allowing any amount of
concurrent, on-demand testing, greatly improving productivity. I then
write an internal I/O Reliability Enhancement research report and
happen to mention MVS 15min MTBF, bringing down the wrath of the MVS
organization on my head.
Fall 1980 Jim Gray leaves for Tandem, palming off some amount of stuff
on me. At Tandem, Jim does studies of service uptime, finding the
hardware was starting to get more reliable and downtime was starting
to shift to things like environmental (power, floods, earthquakes,
etc), people mistakes, etc.
1988, branch office asks if I could help LLNL standardize some serial
stuff they were working with, which quickly becomes fibre-channel
standard ("FCS", including some stuff I had done in 1980, initially
1gbit transfer, full-duplex, aggregate 200mbyte/sec transfer). Then
IBM POK mainframe releases their serial stuff (when it is already
obsolete), "ESCON", 10mbyte/sec. Then some IBM POK engineers becomes
involved with FCS and define a heavy-weight protocol that radically
reduces throughput eventually released as "FICON".
Latest public benchmark I've found is 2010 z196 "Peak I/O" getting 2M
IOPS using 104 FICON (20K IOPS/FICON). About the same time a native
FCS is announced for E5-2600 server blades getting over million IOPS
(two such FCS higher throughput than 104 FICON running over FCS). Note
IBM docs has SAPs (system assist processors that do actual I/O) CPU be
kept to 70% ... or 1.5M IOPS ... also no CKD DASD have been made for
decades (just simulated with industry standard fixed-block devices).
This "Evolution of the System z Channel" (SHARE 2011 presentation at
wayback machine)
https://web.archive.org/web/20170829213251/https://share.confex.com/share/117/webprogram/Handout/Session9931/9934pdhj%20v0.pdf
talks about after 2010 z196 "Peak I/O" benchmark and "zHPF" that
increased throughput from 20K IOPS/FICON to 52K IOPS/FICON then to 92K
IOPS/FICON (aka from 1/50th to almost 1/10th the native FCS announced
in 2010).
Also 1988 HA/6000 is approved, originally for NYTimes to move their
newspaper system ("ATEX") off DEC VAXCluster to RS/6000. I rename it
HA/CMP
https://en.wikipedia.org/wiki/IBM_High_Availability_Cluster_Multiprocessing
when I start doing technical/scientific cluster scale-up with national
labs (LANL, LLNL, NCAR, etc) and commercial cluster scale-up with
RDBMS vendors (Oracle, Sybase, Ingres, Informix) that had VAXCluster
support in same source base with Unix (I do distributed lock manager
with VAXCluster semantics to ease the RDBMS ports).
The IBM Toronto lab was doing a "portable" RDBMS ("SHELBY")
implementation for OS2 ... but it was quite simple (including no
cluster support).
IBM S/88 (relogo'ed Stratus) Product Administrator started taking us
around to their customers and also had me write a section for the
corporate continuous availability strategy document (it gets pulled
when both AS400/Rochester and mainframe/POK complain they couldn't
meet requirements). Had coined disaster survivability and
geographic survivability (as counter to disaster/recovery) when out
marketing HA/CMP.
One of the first HA/CMP customer installs was new Indian Reservation
Casino in Connecticut, was suppose to have week of testing before
opening ... but after 24hrs, they decided to open the doors (based on
projected revenue; at the time was largest in the US, still one of the
largest in the country)
https://en.wikipedia.org/wiki/Foxwoods_Resort_Casino#Debt_default
Early Jan92, there was HA/CMP meeting with Oracle CEO and IBM/AWD
executive Hester tells Ellison that we would have 16-system clusters
by mid92 and 128-system clusters by ye92. Mid-jan92, I update FSD on
HA/CMP work with national labs and FSD decides to go with HA/CMP for
federal supercomputers. By end of Jan92, we are told that cluster
scale-up is being transferred to Kingston for announce as IBM
Supercomputer (SP, technical/scientific *ONLY*) and we aren't allowed
to work with anything that has more than four systems (we leave IBM a
few months later). A couple weeks later, 17feb1992, Computerworld news
... IBM establishes laboratory to develop parallel systems (pg8)
https://archive.org/details/sim_computerworld_1992-02-17_26_7
Some speculation that commercial HA/CMP would have eaten IBM mainframe
in commercial market ... 1993 benchmark (number of program iterations
compared to industry MIPS reference platform):
ES/9000-982 : 8CPU 408MIPS, (51MIPS/CPU)
RS6000/990 (RIOS chipset) : 1-CPU: 126MIPS
The executive we had reported to goes over to head up Somerset/AIM
(Apple, IBM, Motorola) to do single chip 801, uses Motorola 88K
cache&bus enabling multiprocessor configurations.
In the 70s, my (future) wife had been in the Gburg JES group and one
of the catchers for ASP/JES3 and co-author of JESUS (all the features
of JES2&JES3 that the respective customers couldn't live w/o). She is
then con'ed into going to POK to be in charge of loosely-coupled
architecture (Peer-Coupled Shared Data). She didn't remain long, 1)
periodic battles with communication group trying to force her into
using SNA/VTAM for loosely-coupled operation and 2) little uptake
(until much later with SYSPLEX and Parallel SYSPLEX) except for IMS
"hot-standby" (she has story asking Vern who he would ask permissions
to do "hot-standby" and he says nobody, he would just tell them when
it was all done).
https://www.vcwatts.org/ibm_story.html
Before and after leaving IBM, the person at NYFED responsible for
FEDWIRE liked us to stop by and talk technology. He attributed FEDWIRE
100% (IMS hot-standby) availability for over a decade to
1) automated operator and 2) three redundant hot-standby IMS
systems at two physically separated locations.
SMP, tightly-coupled, shared memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
original sql/relational System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
getting to play disk engineer posts
https://www.garlic.com/~lynn/subtopic.html#disk
FCS and/or FICON posts
https://www.garlic.com/~lynn/submisc.html#ficon
HA/CMP posts
https://www.garlic.com/~lynn/subtopic.html#hacmp
HASP/JES2, ASP/JES3, and/or NJI/NJE posts
https://www.garlic.com/~lynn/submain.html#hasp
Peer-Coupled Shared Data architecture posts
https://www.garlic.com/~lynn/submain.html#shareddata
some posts mentioning IMS hot-standby & FEDWIRE
https://www.garlic.com/~lynn/2026.html#44 IBM Loosely-coupled and Hot Standby
https://www.garlic.com/~lynn/2024c.html#105 Financial/ATM Processing
https://www.garlic.com/~lynn/2023f.html#72 Vintage RS/6000 Mainframe
https://www.garlic.com/~lynn/2022c.html#20 Telum & z16
https://www.garlic.com/~lynn/2021i.html#19 A brief overview of IBM's new 7 nm Telum mainframe CPU
https://www.garlic.com/~lynn/2021c.html#39 WA State frets about Boeing brain drain, but it's already happening
https://www.garlic.com/~lynn/2021b.html#72 IMS Stories
https://www.garlic.com/~lynn/2012d.html#6 NASA unplugs their last mainframe
https://www.garlic.com/~lynn/2012.html#67 Has anyone successfully migrated off mainframes?
https://www.garlic.com/~lynn/2011n.html#3 Banks Face Ongoing Cyber Threats
https://www.garlic.com/~lynn/2009r.html#20 70 Years of ATM Innovation
https://www.garlic.com/~lynn/2008p.html#5 Privacy, Identity theft, account fraud
https://www.garlic.com/~lynn/2008o.html#76 Blinkenlights
--
virtualization experience starting Jan1968, online at home since Mar1970
history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
Self-hosting and the 6502)
Newsgroups: alt.folklore.computers
Date: Sat, 12 Sep 2026 14:12:05 -1000
Kragen Javier Sitaker <kragen@canonical.org> writes:
trivia: old archived post with decade of VAX/VMS numbers ... VM/4341s
sold in approx. same numbers in single or small unit numbers ... big
difference were the large orders for hundreds of VM/4341 at a time.
https://www.garlic.com/~lynn/2002f.html#0
Sad to say, garlic.com seems to have broken your link here.
https://web.archive.org/web/20260310072508/https://www.garlic.com/~lynn/2002f.html#0
works for now. Your post archive links are extremely valuable, and I
hope you can get your archive back on the live web — and maybe license
it under a Creative Commons license so that it's clearly legal for
people to archive.
... PCs & PC browser seem to work ... it seems to be mostly cellphone
and tablet for which it has disappeared.
re:
https://www.garlic.com/~lynn/
--
virtualization experience starting Jan1968, online at home since Mar1970
history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
Self-hosting and the 6502)
Newsgroups: alt.folklore.computers
Date: Sat, 12 Sep 2026 14:13:35 -1000
Jeroen Belleman <jeroen@nospam.please> writes:
Where did Lynn's article come from?
the CERN/MVS/VM370 comparison reference came from being there
https://www.garlic.com/~lynn/2026c.html#36
--
virtualization experience starting Jan1968, online at home since Mar1970
history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
Self-hosting and the 6502)
Newsgroups: alt.folklore.computers
Date: Sat, 12 Sep 2026 14:47:44 -1000
Jeroen Belleman <jeroen@nospam.please> writes:
When I started at CERN in 1981, the principal computer
system was an IBM 360. We used WYLBUR for the editing
environment, with exec files for simple things and batch
processing with JCL for the more serious work.
It was clumsy, but I didn't know any better at the time.
trivia: WYLBUR original came from stanford's virtual memory
system done for 360/67 ... Univ of Michigan had done MTS
https://en.wikipedia.org/wiki/Michigan_Terminal_System
and Stanford had done ORVYL
https://en.wikipedia.org/wiki/ORVYL_and_WYLBUR
... later WYLBUR was ported to MVS
Some of the MIT CTSS/7094 people had gone to 5th flr to do
multics. Others went to the IBM Scientific Center on the 4th flr and did
virtual machines (CP40/CMS morphs into CP67/CMS before morphing
into VM370/CMS). CMS SCRIPT (redone CTSS RUNOFF) was done for CP67/CMS in
1968. Then some CSC co-workers invented GML in 1969 and GML tag
processing was added to SCRIPT. After a decade, GML morphs into ISO
standard SGML and after another decade morphs into HTML at CERN.
Also account by one of the GML inventors ... original hired to promote
CP67-based scientific center wide-area network
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for the
project, something I was allowed to do part-time because of my knowledge
of the user requirements. My real job was to encourage the staffs of the
various scientific centers to make use of the CP-67-based Wide Area
Network that was centered in Cambridge.
... snip ...
CSC wide-area network morphs into the internal corporate network (larger
than arpanet/internet from the beginning until sometime mid/late 80s
... about the time it was forced to convert to SNA/VTAM). Technology was
also used for the corporate sponsored univ BITNET
https://en.wikipedia.org/wiki/BITNET
Edson (responsible for science center wide area entwork)
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which paved
the way to the Internet as we know it today.
...
SJMerc article about Edson (he passed aug2020) and "IBM'S MISSED
OPPORTUNITY WITH THE INTERNET" (gone behind paywall but lives free at
wayback machine)
https://web.archive.org/web/20000124004147/http://www1.sjmercury.com/svtech/columns/gillmor/docs/dg092499.htm
Also from wayback machine, some additional (IBM missed internet)
references from Ed's website
https://web.archive.org/web/20000115185349/http://www.edh.net/bungle.htm
Ed and I transfer out to SJR on the west coast in 1977 ... and I work
with Jim Gray and Vera Watson on original SQL/relational, System/R. Was
able to do technology transfer ("under the radar while company was
pre-occupied with next great new DBMS "EAGLE") to Endicott for SQL/DS.
Then when "EAGLE" implodes, there was request for how fast could
System/R be ported to MVS, eventually released as DB2 (originally for
decision support *ONLY*).
CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
GML, SGML, HTML posts
https://www.garlic.com/~lynn/submain.html#sgml
internal network
https://www.garlic.com/~lynn/subnetwork.html#internalnet
original SQL/Relational, System/R posts
https://www.garlic.com/~lynn/submain.html#systemr
Also TYMSHARE provides the IBM Mainframe user group, "SHARE" with its
VM370/CMS-based online computer conferencing system for free in AUG1976.
I cut a deal with TYMSHARE to get a monthly tape dump of all VMSHARE
files for putting up on internal network and systems (biggest problem
were the lawyers over exposing internal employees to unfiltered customer
information). archives here:
http://vm.marist.edu/~vmshare
--
virtualization experience starting Jan1968, online at home since Mar1970
position-independent code (was Re: CMS, Self-hosting and the
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: position-independent code (was Re: CMS, Self-hosting and the 6502)
Newsgroups: alt.folklore.computers, alt.sys.pdp8
Date: Sat, 12 Sep 2026 15:12:56 -1000
Kragen Javier Sitaker <kragen@canonical.org> writes:
I've never programmed the PDP-8, but, as I understand it, its 7-bit
address field in the 12-bit instruction format effectively divides the
4096-word address space into 128-word "pages". An 8th bit specifies
which page the address field is interpreted relative to: either a
6502-like zero page, or the current page. So, in effect, you have two
5-bit "base registers": one hardwired to 00000, and the other one that's
the high 5 bits of the program counter. (Later PDP-8s extended the
address bus to the left and needed additional registers to select which
12-bit memory space you were running in.)
OS/360 compilers generated TXT decks that had RLD (relocation directory)
unresolved addresses for other programs that were being loaded in one
large group (their "ESD" entry, the ID of program being referenced and
displacement within the program). The "loader" would resolve the "fixed"
desired addresses (from combination of ESD and displacement) before execution.
CP67/CMS adopted OS/360 compilers (and later VM370/CMS the OS/370
compilers).
By comparison TSS/360 kept an external map for each virtual address
space .... where its programs were loaded. Issue was initially virtual
address space was 16 1mbyte virtual address segments. Somebody might 1st
load the fortran compiler into its segment one and somebody else might
have first loaded something else into segment one and then also wanted
to load the ("same") fortran compiler image into its virtual address
space.
The OS/360 paradigm wouldn't support it ... it needed either
1) a different copy of the fortran compiler image loaded, if it was to
run at different address or
2) had a predefined location for every application in the system (in
order to have virtual address space to have a single shared application
image). With only 16 shared segments, would only be able to have 16
locations for predefined "shared image" application copies.
position-independent code
https://www.garlic.com/~lynn/submain.html#adcon
--
virtualization experience starting Jan1968, online at home since Mar1970
position-independent code (was Re: CMS, Self-hosting and the 6502)
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: position-independent code (was Re: CMS, Self-hosting and the 6502)
Newsgroups: alt.folklore.computers
Date: Mon, 14 Sep 2026 14:41:47 -1000
James Dow Allen <user4353@newsgrouper.org.invalid> writes:
It was half a century ago that I handled physical TXT card decks.
All my programs were standalone: Slap a "six-card loader" at the front of
the TXT deck, and IPL.
The source code for my programs always started with
USING *,0
Did that suppress any RLDs? Don't remember. The six-card loader ignored
RLDs anyway. Actually -- never one to overlook a pointless micro-
optimization -- I "optimized" down to a three-card loader, good enough if
the TXT deck had no REP (hexadecimal patch) cards.
re:
https://www.garlic.com/~lynn/2026c.html#42 position-independent code (was Re: CMS, Self-hosting and the 6502)
I had taken two credit hr intro to Fortran/Computers. At the end of
semester was hired to rewrite 1401 MPIO in 360 assembler for 360/30
(64kbyte memory and 2311s). Univ was getting 360/67 for tss/360
replacing 709/1401 (709 tape->tape, 1401 unit record frontend) and got
360/30 temporarily replacing 1401 pending arrival of 360/67.
The univ. shutdown the datacenter on weekends and I would have the place
dedicated. They gave a bunch of hardware and software manuals and I got
to design and implement my own monitor, device drivers, interrupt
handlers, storage management, error recovery, etc ... and within a few
weeks I had a 2000 card assembler program. I didn't know about rep
cards, so learned to read TXT deck cards, duplicate cards, and
multi-punch hex patches.
Univ. 360/67 arrived within year of taking intro class and I was hired
fulltime responsible for OS/360 (768kbte memory and 2314s; TSS/360 not
coming to production, so ran as 360/65). 709 did student fortran in less
than second. 360/67 (as 360/65) os/360 (MFT9.5) took over a minute, I
install HASP cutting time in half. MFT11, I start redoing SYSGEN STAGE2
to carefully place datasets and PDS members to optimize arm seek and
(PDS directory) multi-track search, cutting another 2/3rds to
12.9secs. OS/360 never got better than 709 until I install Univ Waterloo
WATFOR; as 360/65, WATFOR clocked at 20,000 statements/min (333
statements/sec), student fortran tended to run 30-60 statements.
Then IBM CSC came out to install (virtual machine, precursor to vm370)
CP/67 (3rd after CSC itself and MIT Lincoln Labs) and I mostly get to
play with it during my weekend 48hr window. I then spend a few months
rewriting pathlengths for running OS/360 in virtual machine. Source was
in OS/360 files which had to be modified, assembled, mark txt deck and
merge it into a tray of txt decks (that was CP67 executable). BPS loader
IPL the tray of TXT decks which then writes core image to disk, for
ipl'ing the system.
Bare machine test ran 322secs ... initially 856secs (CP67 CPU
534secs). After a few months I had CP67 CPU down from 534secs to
113secs. I then start rewriting the dispatcher/scheduler , (dynamic
adaptive resource manager/default fair share scheduling policy), paging,
adding ordered seek queuing (from FIFO) and mutli-page transfer channel
programs (from FIFO and optimized for transfers/revolution, getting 2301
paging drum from 70-80 4k transfers/sec to channel transfer peak of
270). Six months after univ initial CP/67 install, CSC was giving one
week class in LA. I arrive on Sunday afternoon and asked to teach the
class, it turns out that the people that were going to teach it had
resigned the Friday before to join one of the 60s CSC CP67 online
commercial spin-offs (one offered Mathematica RAMIS, when RAMIS was also
offered to TYMSHARE, they did Nomad, then some former Mathematica do
FOCUS at Information Builders)
Before I graduate, I'm hired fulltime into small group in Boeing CFO
office to help with formation of Boeing Computer Services (consolidate
all dataprocessing into independent business unit). I think Renton,
largest datacenter in the world, 360/65s arriving faster than they could
be installed, boxes constantly staged in hallways around machine room
(did have lone 360/75, when doing classified work; black rope around
area, guards at corners, heavy black felt covering console lights and
visible 1403 printer areas). Lots of politics between Renton director
and CFO, who only had a 360/30 up at Boeing Field for payroll, although
they enlarge the room to install 360/67 for me to play with when I
wasn't doing other stuff. I had mid-level management badge so could park
in lots closer to the bldgs.
747-3 was flying skies of Seattle getting FAA flt certification. Tours
of mock-up of 747 cabin just south of Boeing field would claim 747s
carried so many passengers, they would never be serviced with fewer than
four jetways. Both Boeing and IBM told stories about 360 announcement
day, Boeing submitted an order making the IBM marketing rep, highest
paid IBM employee that year. Sales were still straight commission; IBM
shifts to quota the following year.
At Boeing I modified CP67 to support portions as pageable
kernel. Problem was adding more CSECTs (dividing pageable kernel into 4k
segments). Problem was I ran into BPS loader had 255 csect limit ...
and I was constantly doing hacks to keep the pageable kernel within 255
limit.
When I graduated, I joined CSC (instead of staying with Boeing CFO) and
one of my hobbies was enhanced production operating systems for internal
datacenters (one of the first was the branch office online
sales&marketing support HONE was one of the 1st & long time customers).
CSC had several card drawer cabinets up on the 9th flr attic ... and
prowling through them found source for BPS loader, and modified it to
double the maximem csects ... greatly simplifying pageable kernel
(wasn't released to customers, but picked up for VM370 follow-on).
Co-worker was responsible for CP67-based scientific centers wide-area
network. He had three "PUNCH" statements at front of the network
assembler source ... that punched hex for a 3card loader. This morphs
into the internal corporate network (larger than arpanet/internet from
the beginning until sometime mid/late 80s, about the time it was forced
to convert to SNA/VTAM).
Some of the MIT CTSS/7094 had gone to the 5th flr to do MULTICS and
others went to the scientific center on the 4th flr and done virtual
machines, networking, invented GML 1969 (morphs into SGML a decade
later, after another decade morphs into HTML at CERN), etc
Looking at what they were doing for MULTICS on the 5th flr, I created
page-mapped filesystem for CP67/CMS ... want to be able to just map
virtual memory addresses to pages on disk (w/o having to preload them
for the loader could modify the addresses) and support location
independent R/O shared (executable) files ... which I later migrate to
VM370/CMS.
For VM370R3, there is a very small subset released as DCSS, extending
shared segment to more than kernel supported with more than just the IPL
command.
IBM CSC posts
https://www.garlic.com/~lynn/subtopic.html#545tech
internal CP67L, CSC/VM, SJR/VM posts
https://www.garlic.com/~lynn/submisc.html#cscvm
HONE posts
https://www.garlic.com/~lynn/subtopic.html#hone
paged-mapped filesystem
https://www.garlic.com/~lynn/submain.html#mmap
--
virtualization experience starting Jan1968, online at home since Mar1970
history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
Newsgroups: alt.folklore.computers
Date: Mon, 14 Sep 2026 15:06:05 -1000
Kragen Javier Sitaker <kragen@canonical.org> writes:
This unlikely to be related to which browser you're using, but in case
it is, I'm using Firefox 128.3.1esr for Linux.
re:
https://www.garlic.com/~lynn/2026c.html#39 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
https://www.garlic.com/~lynn/2026c.html#40 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
https://www.garlic.com/~lynn/2026c.html#41 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
firefox/nightly158 w/fedora44 ... it gives me
https://www.garlic.com/~lynn/ and https://mycloud.garlic.com/~lynn
... but neither worked with tablet or cellphone.
just retried with cellphone firefox and mycloud.garlic.com worked but
not www.garlic.com. I've had discussions with garlic tech guy and he
says he had been working on getting something fixed (after recent site
revamp).
just double checked wayback machine (archive.org) which had
www.garlic.com/~lynn up until about may2026 but nothing since ... and
nothing at all for mycloud.garlic.com.
--
virtualization experience starting Jan1968, online at home since Mar1970
history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
From: Lynn Wheeler <lynn@garlic.com>
Subject: Re: history of ASCII and IBM ratfucking, and the VAX (was Re: CMS, Self-hosting and the 6502)
Newsgroups: alt.folklore.computers
Date: Mon, 14 Sep 2026 15:11:58 -1000
Lars Poulsen <lars@beagle-ears.com> writes:
I wonder why there were so many CRJE systems with almost identical features:
• WYLBUR (from Stanford)
• WITS (from Waterloo)
• CRJE (IBM's official option for MVT and MVS)
... any that I have forgotten?
Of course, there was also TSO, but it was a dog, performance wise, so
none of the education places I knew would use it. I expect that the
IBM develoment sites used it to develop MVS code?
re:
https://www.garlic.com/~lynn/2026c.html#39 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
https://www.garlic.com/~lynn/2026c.html#40 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
https://www.garlic.com/~lynn/2026c.html#41 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
https://www.garlic.com/~lynn/2026c.html#41 history of ASCII and IBM ratfucking, and the VAX (was Re: CMS,
CP/67 had arrived with 1052&2741 terminal support and auto-terminal
ident, capable switching terminal type scanner type for each
port. Univ. also had TTY33&35 and I add ASCII support integrated with
auto-terminal type. I then want to have single dial-in number (hunt
group) for all terminals. Didn't quite work, IBM had hard-wired line
speed ... so we start a clone terminal controller. Build a IBM channel
interface board for Interdata/3 programmed to emulate IBM controller
with the addition for auto-baud. Then upgraded with Interdata/4 for
channel interface and cluster of Interdata/3s for port
interfaces. Interdata (and later Perkin-Elmer) sell them as clone
controllers.
https://en.wikipedia.org/wiki/Interdata
https://en.wikipedia.org/wiki/Perkin-Elmer#Computer_Systems_Division
... and four of us are written up responsible for (some part of) IBM
clone controller business. Other trivia, when tty/ascii port scanner for
360/67, it came in "Heathkit" box.
Before leaving for Boeing. I modified HASP for MVT18, with CRJE
implementation, including editor that implemented CMS edit syntax
(environments were so different that couldn't reuse any code).
clone controller posts
https://www.garlic.com/~lynn/submain.html#360pcm
HASP/JES2, ASP/JES3, NJI/NJE posts
https://www.garlic.com/~lynn/submain.html#hasp
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM Networking
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM Networking
Date: 15 Sep, 2026
Blog: Facebook
co-worker at the ibm cambridge scientific center was responsible for
the CP67-based scientific center wide-area network. one of the 1969
inventors of GML (after a decade morphs into SGML, after another
decade morphs into HTML at CERN) originally hired to promote the
network
https://web.archive.org/web/20230402212558/http://www.sgmlsource.com/history/jasis.htm
Actually, the law office application was the original motivation for
the project, something I was allowed to do part-time because of my
knowledge of the user requirements. My real job was to encourage the
staffs of the various scientific centers to make use of the
CP-67-based Wide Area Network that was centered in Cambridge.
... snip ...
morphs into the IBM corporate internal network (larger than
arpanet/internet from the beginning until sometime mid/late 80s
... about the time it was forced to convert to SNA/VTAM), technology
also used for the corporate sponsored univ BITNET (EARN in Europe)
https://en.wikipedia.org/wiki/BITNET
Edson was responsible for CP67 wide-area network
https://en.wikipedia.org/wiki/Edson_Hendricks
In June 1975, MIT Professor Jerry Saltzer accompanied Hendricks to
DARPA, where Hendricks described his innovations to the principal
scientist, Dr. Vinton Cerf. Later that year in September 15-19 of 75,
Cerf and Hendricks were the only two delegates from the United States,
to attend a workshop on Data Communications at the International
Institute for Applied Systems Analysis, 2361 Laxenburg Austria where
again, Hendricks spoke publicly about his innovative design which
paved the way to the Internet as we know it today.
...
SJMerc article about Edson (he passed aug2020) and "IBM'S MISSED
OPPORTUNITY WITH THE INTERNET" (gone behind paywall but lives free at
wayback machine)
https://web.archive.org/web/20000124004147/http://www1.sjmercury.com/svtech/columns/gillmor/docs/dg092499.htm
Also from wayback machine, some additional (IBM missed internet)
references from Ed's website
https://web.archive.org/web/20000115185349/http://www.edh.net/bungle.htm
science center posts
https://www.garlic.com/~lynn/subtopic.html#545tech
internal network posts
https://www.garlic.com/~lynn/subnetwork.html#internalnet
BITNET (& EARN) posts
https://www.garlic.com/~lynn/subnetwork.html#bitnet
internet posts
https://www.garlic.com/~lynn/subnetwork.html#internet
some specific posts this year
https://www.garlic.com/~lynn/2026c.html#33 Internet Lore
https://www.garlic.com/~lynn/2026c.html#29 HSDT
https://www.garlic.com/~lynn/2026c.html#21 CMSBACK, ADSM, TSM
https://www.garlic.com/~lynn/2026c.html#13 Remember when discussions were worth reading?
https://www.garlic.com/~lynn/2026c.html#9 IBM Mainframe Demise
https://www.garlic.com/~lynn/2026b.html#107 Network History
https://www.garlic.com/~lynn/2026b.html#91 IBM CEOs and IBM Downfall
https://www.garlic.com/~lynn/2026b.html#87 ARPANET and Internet
https://www.garlic.com/~lynn/2026b.html#81 Token-Ring
https://www.garlic.com/~lynn/2026b.html#64 More IBM Downturn and Downfall
https://www.garlic.com/~lynn/2026b.html#37 IBM Internal Network and BITNET
https://www.garlic.com/~lynn/2026b.html#36 IBM RS/6000 HA/CMP
https://www.garlic.com/~lynn/2026b.html#28 How We Put It Together
https://www.garlic.com/~lynn/2026b.html#27 IBM Mainframe
https://www.garlic.com/~lynn/2026b.html#26 IBM CICS
https://www.garlic.com/~lynn/2026.html#41 Ethernet and IBM Token-Ring
https://www.garlic.com/~lynn/2026.html#32 IBM SNA
--
virtualization experience starting Jan1968, online at home since Mar1970
IBM 370 Virtual Memory
From: Lynn Wheeler <lynn@garlic.com>
Subject: IBM 370 Virtual Memory
Date: 15 Sep, 2026
Blog: Facebook
Early last decade I was asked to track down executive decision to add
virtual memory to all 370s ... and found staff to the
executive. Basically MVT storage management was so bad, regions had to
be specified four times larger than used and typical 1mbyte 370/165
would only run four regions concurrently, insufficient to keep system
busy and justified. Going to 16mbyte virtual memory allowed concurrent
regions to be increased by factor of four times (capped at 15 because
of 4bit storage protect keys), with little or no paging.
I would drop in on Ludlow periodically who was doing the initial
implementation of MVT->VS2/SVS (sort of like running MVT in CP/67
16mbyte virtual macine) on 360/67 (until engineering 370 appeared with
virtual memory). He had a little bit of code for the 16mbyte virtual
address space and some simple paging. Biggest task was channel
programs passed to EXCP/SVC0 now had virtual addresses and channels
required real addresses and he borrows CP67's CCWTRANS for integrating
into EXCP (creating channel program copies, replacing virtual with
real addresses).
Initial virtual memory CP67->VM370, DOS->DOS/VS, MFT->VS1,
MVT->VS2/SVS morphing into VS2/MVS ... aka MVS gave each region its
own 16mbyte virtual address space (for protection in lieu of storage
protect key caping regions at 15 concurrent).
Downside of VS2/MVS was MVT APIs were heavily pointer passing and so
needed to map an 8mbyte image of the MVS kernel into every address
space (leaving 8mbyes). Then every subsystem went to their own 16mbyte
virtual address space and for regions passing info to/from subsystem
calls, mapped a one mbyte "Common Segment Area" (CSA) into every
virtual address space, leaving 7mbytes. Common area requirement
somewhat proportional to number of concurrent regions and subsystems
... and by 3033 common area had exploded to 5-6mbytes as Common System
Area (still "CSA", "leaving 3-2mbytes and threatening to become 8mbyte
CSA, leaving zero) .... part of the desperate need getting to MVS/XA.
"Future System" was spawned after start of 370 virtual memory effort,
was completely different from 370 and was to completely replace 370
(lack of new 370 products during FS is credited with giving the clone
370 makers, their market foothold). When FS implodes there is mad rush
to get stuff back into the 370 product pipelines, including kicking
off the quick&dirty 3033&3081 in parallel.
http://www.jfsowa.com/computer/memo125.htm
https://en.wikipedia.org/wiki/IBM_Future_Systems_project
https://people.computing.clemson.edu/~mark/fs.html
CERN had done a 1974 SHARE presentation of analysis comparing
VM370/CMS with MVS/TSO, was freely available ... except inside IBM. It
was stamped "IBM Confidential - Restricted" (available on "need to
know" only). Likely part of the motivation for head of POK getting
corporate to kill the VM370 product, shutdown the development group
and transfer all the people to POK for MVS/XA. Endicott eventually
managed to acquire the VM370 product responsibility for the mid-range,
but had to recreate a development group from scratch. They weren't
planning on telling the VM370 group about shutdown and move to
minimize the number that might escape into the Boston area. It managed
to leak early and several managed to escape (it was in early infancy
of DEC VAX/VMS and joke was that head of POK was major contributor to
VMS). There was witch hunt for the leaker, fortunately for me, nobody
gave them up.
trivia: prior to graduation and joining IBM, I was hired fulltime into
small group in the Boeing CFO office to help with the formation of
Boeing Computer Services (consolidate all dataprocessing in an
independent business unit). I think Renton datacenter possibly largest
in world. IBMers and Boeing told story about S/360 announcement day,
Boeing walks into marketing rep's office and gives him order making
him the highest paid IBM employee that year (straight commission,
following year IBM deploys "quota"). When I graduate, I join IBM
science center (instead of staying with Boeing CFO).
Boeing Huntsville had ordered a 2-CPU 360/67 for TSS/360 with lots of
graphic 2250s for CAD/CAM, however TSS/360 never really came to
production. Boeing then run it as two 360/65 MVT13 systems. It ran
into the MVT storage management problems early on ... and done some
really simple mods to MVT13 to run in virtual memory mode as partial
work around.
more trivia: 23Jun1969 unbundling announce started to charge for
(application) software (but managed to make the case that kernel
software was still free). Then with the rise of clone 370 makers
(during FS) and then the FS implosion, it was decided to start
charging for kernel (system) software. Initially it would be kernel
add-ons planning on evolving to charging for all kernel software in
the 80s (followed by the OCO-wars). When I left Boeing CFO and joined
IBM scientific center, one of my hobbies was enhanced production
operating systems for internal datacenters (branch office online
sales&marketing support HONE was one of my 1st and long-time
customers). A bunch of my internal enhancements was selected to be the
initial guinea pig for kernel software charging. Before release, a
corporate specialist (steeped in MVS) evaluated and found no manual
tuning knobs. He said he wouldn't sign-off because everybody knew that
manual tuning knobs was the state of the art. I created a set of
manual tuning knobs packaged as "DMKSRM" (to ridicule MVS) and the
dynamic adaptive implementation was packaged as "DMKSTP" (after TV
commercial ... "DMK" was the code prefix for VM370 kernel). The joke
was that the dynamic adaptive DMKSTP features had more "degrees of
freedom" and could offset any DMKSRM manual setting.
SMP, tightly-coupled, shared-memory multiprocessor posts
https://www.garlic.com/~lynn/subtopic.html#smp
Future System posts
https://www.garlic.com/~lynn/submain.html#futuresys
IBM 23jun1969 unbundling posts
https://www.garlic.com/~lynn/submain.html#unbundle
dynamic adaptive resource management, "wheeler" scheduler posts
https://www.garlic.com/~lynn/subtopic.html#fairshare
--
virtualization experience starting Jan1968, online at home since Mar1970
--
previous, index - home