Post by David LesherNo, not TSO... TSS.
I used it at NASA-LeRC, and know Ames used it too.
I just wonder if anyone here had.
univ had 709 with 1401 for unit record front end when I took two
semester hour intro to computers/fortran. the univ. was talked into
replacing 709/1401 with 360/67 for tss/360. at the end of the semester
the 1401 was (temporarily) replaced with 360/30 (as part of transition
to 360/67) and I got programming job to redo 1401 MPIO (tape<->unit
record) on 360/30 (they could have just ran 1401 MPIO on 360/30 in 1401
hardware emulation mode, but they apparently wanted to get some 360
experience). I got to design & implement my own monitor, commands,
device drivers, interrupt handlers, error recovery, storage management,
etc. Univ. shutdown datacenter on weekends and I (mostly) had datacenter
all to myself (although 48hrs w/o sleep could make monday morning
classes difficult). Within about ten weeks, I had 2000 card assembler
program that took 30mins to assemble under os/360 on 360/30 ... for
stand-alone version loaded with BPS loader ... or 60mins for version
that ran under OS/360 (DCB macros took 5-6 mins elapsed time each).
768kbyte 360/67 then replaced 709 & 360/30 ... and mostly ran os/360
... tss/360 never quite coming to product fruition. Few months later, I
was hired fulltime to be response of os/360 system. On weekends I
sometimes had to share the machine with IBM SE who was playing with
TSS/360.
Last week Jan1968, three people from the IBM science center came out and
installed CP67/CMS. The IBM SE and I put together a simulated
edit-compile-load-execute benchmark that I ran with 35 simulated users
under CP67 and he ran four simuulated users under TSS/360; the CP67 with
35 users had better interactive response and throughput thatn TSS/360
with four users. TSS/360 was enormous CPU and real storage hog ... IBM
did some TSS/360 1mbyte, single processor benchmarks compared to 2mbyte,
two processor benchmarks which got 3.8 times the throughput of the
single processor. IBM tried to spin that TSS/360 was so sophisticated
that its algorithms got nearly four times the thoughput with just twice
the hardware ... when realisticly TSS/360 needed over mbyte of real
storage just to handle the kernel before it could start getting any user
work done.
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 to
bettern monitize the investment, even offering services to non-Boeing
entities). I thought Renton was possibly largest datacenter in the
world, something like $200M-$300M in 360 gear (60s dollar), 360/65s were
arriving faster than they could be installed, boxes constantly staged in
the hallways around the machine room. They get me a 1mbyte single
process 360/67 to play with and they bring up the dual processor 360/67
to Seattle from Boeing Huntsville (that had been configured as two
360/65 running OS/360).
When I graduate, instead of staying at Boeing, I join the cambridge
science center. I do a page-mapped filesystem for CP67/CMS (later ported
to VM370/CMS) that never ships to customers but I get it deployed at a
lot of installations (one of my hobbies after joining IBM was production
operating systems for internal datacenters, including the online
world-wide sales&marketing HONE systems). In moderately heavy filesystem
benchmarks get three times the throughput of standard CMS filesystem
... I would say I learned what *NOT* to do for paged-mapped filesystem
from TSS/360.
TSS/370 finally starts to come into production quality ... a lot of
performance work was done after they cut the organization from 1200
people to 20 people and they start getting 370/168 machines with 4mbytes
of memory (enough to start running applications w/o heavy page
thrashing). I had interactions off and on with the group through the
mid-80s.
--
virtualization experience starting Jan1968, online at home since Mar1970