[email protected] (John Eells) writes:
> It's certainly true that running the first large MVS system required a
> significant number of people (operators, production control, system
> programmers, etc.).  However, the second through *n*th had far lesser
> incremental cost.  I was a sysprog during this period and we supported
> 30-34 MVS systems (depending on the year) plus several VM systems with
> perhaps 3 people per MVS system overall.
>
> I know of at least one customer who ran a one-person system
> programming shop at the time, too.
>
> Also, because a substantial fraction of system programmer time was
> devoted to debugging at the time, I rather suspect the average ratio
> (whatever that might be) has improved markedly.

re:
http://www.garlic.com/~lynn/2014g.html#83 Costs of core

CP67 & VM370 took some time to get to unattended operation.

Big first push was in the 60s with CP67 and being able to offer 7x24
online availability ... system costs were recovered by billing and
leaving system up 7x24 when (initially) there was very low useage
... was big transition ... part of it was reducing total cost of
operation as much as possible ... automatic ipl w/o human intervention,
lots of other stuff w/o human interventiona. 

This was also in the period that machines were leased/rented with
charges based on the system meter (which ran whenever processor and/or
any channel was busy). An early gimick was special CCW so that things
were reading for incoming characters ... but the channel wasn't
classified as busy (so system meter would stop if there wasn't actually
any activity). One of the characteristics of the system meter was all
activity had to be quiet for at least 400ms before it actually stopped.
Note that long after systems had converted from lease to sales ...  and
system meter was important ... MVS still had an internal system task
that woke up every 400ms (to make sure that system meter never stopped).

There was a problem as more and more services moved into virtual
machines (common terminology now calls it virtual applicance, but then
was called service virtual machines) ... that even if the system auto
re-ipl'ed w/o human intervention ... there was increasing need to have
the services running virtually automatically come up.

I was doing lots of performance enhancements and turning at the
science center
http://www.garlic.com/~lynn/subtopic.html#545tech

and developed automatic processes including automatically bringing up
running operational virtual machines (automatically re-ipl between each 
benchmark,
even in some cases re-ipl totally different systems). All during the FS
period, 
http://www.garlic.com/~lynn/submain.html#futuresys

I continued to work on 360/370 stuff ... even periodically ridiculing
the FS stuff (which wasn't exactly a career enhancing activity). Then
with failure of FS and mad rush to get stuff back into 370 product
pipelines .... not just 3033&3081 but software ... which contributed to
decision to release in products various pieces of stuff I was doing
during the period. some old email about moving bunch of stuff from cp67
base to vm370 and discussion of various pieces.
http://www.garlic.com/~lynn/2006v.html#email731212
http://www.garlic.com/~lynn/2006w.html#email750102
http://www.garlic.com/~lynn/2006w.html#email750430

above mentions csc/vm ... one of my hobbies was providing enhanced
production operating systems to internal datacenters. Another piece was
"autolog" ... which I had originally used for automated benchmarking
... but internally was increasingly used to automatically bring up
service virtual machines ... and was included in vm370 release3 for that
purpose. There was also a bunch of my integrity stuff picked up for
release3 ... that eliminated all sorts of system failures.

Then it was decided to pickup other pieces for release as the "Resource
Manager" (initially with release3 plc9, aka nineth monthly update). Note
that the 23jun69 unbundling announcement there was start charging for
software, but they managed to make the case that kernel software should
still be free. 
http://www.garlic.com/~lynn/submain.html#unbundle

However, the lack of 370 products during the FS period is credited with
giving the clone processors market foothold. Coming out of the FS
failure, it was decided to start transition to charging for kernel
software ... and my resource manager was selected as the guinea pig as
separately charged for component. Also as part of the release, I had
over 2000 (automated) benchmarks that took 3months elapsed time
... usually carefully selected configurations and workloads to validate
correct operation. some posts
http://www.garlic.com/~lynn/submain.html#benchmark

after I transferred from science center to San Jose research, the
internal distribution csc/vm morphed into sjr/vm. I also got sucked into
doing significant upgrade for the disk engineering and development labs.
They were operating pre-scheduled, stand-alone testing on numerous
mainframe around the clock, 7x24. At one point they had tried MVS for
concurrent testing ... but found it had 15min MTBF in that environment.
I offered to rewrite I/O supervisor to make it bullet proof and never
fail ... so that they could do on-demand concurrent testing
... significantly improving productivity. Afterwards, I would
increasingly be dragged into playing disk engineer
http://www.garlic.com/~lynn/subtopic.html#disk

I was also doing DBMS stuff ... Bank of america was early adopter of the
original sql/relational implementation System/R (that had been done at
vm370 370/145 at san jose research) on 60 vm/4341s. Old email proding me
that I needed to further improve (reduce) the care&feeding for large
number of distributed vm/4341s. old email as Jim was leaving IBM for
tandem, he was palming a bunch of stuff on me, working with BofA,
consulting to the STL IMS group, etc.
http://www.garlic.com/~lynn/2007.html#email801016
earlier email from Jim about BofA (& 60 4341s)
http://www.garlic.com/~lynn/2010.html#email800311
posts mentioning original sql/relational system/r
http://www.garlic.com/~lynn/submain.html#systemr

4300s were selling against DEC VAX machines in the low&mid range market.
they sold about the same number in market involved small numbers of
machines ... big difference was large corporations ordering hundreds of
4300s at a time. old post with a decade of VAX sales sliced&diced
by model, year, US/non-us
http://www.garlic.com/~lynn/2002f.html#0

as can be seen in the above, by the mid-80s, low&mid range market was
starting to shift (to workstations and large PCs).

At the time, MVS in this market wasn't even close for
consideration. SHARE was doing studies/reports comparing the amount of
human effort supporting vm/4341 compared to vax/vms ... recent post with
old email about fortran Q (initially went out as iup opt=3 for FORTHX)
... but also references SLAC having done one of the (for SHARE) VMS/CMS
comparisions
http://www.garlic.com/~lynn/2014g.html#email820322
in this post
http://www.garlic.com/~lynn/2014g.html#71

-- 
virtualization experience starting Jan1968, online at home since Mar1970

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to