6. RPI (Rensselaer Polytechnic Institute)

posted Oct 30, 2010, 6:46 AM by Jeff Ogden   [ updated May 11, 2011, 1:21 PM ]
In 1976 Rensselaer Polytechnic Institute (RPI) in Troy, New York, USA became the sixth member of the MTS Consortium.

Garance Dorsehn posted the following article on Tom Valerio's MTS Wiki in April 2005.

From http://mtswiki.westwood-tech.com/mtswiki-index.php/MTS%20at%20RPI:

MTS comes to RPI

While RPI had been running IBM mainframes, student access was mostly via batch jobs (which is to say "punched cards"). As one might expect, catering to students in engineering and the sciences meant there was a large demand for computing on campus. In 1975 RPI had purchased a used(!!) IBM 360/67 from Rice University, in an attempt to do an inexpensive upgrade from their previous mainframe. That machine was simply not large enough to meet the demand, and the students made it clear that they wanted more significant improvements made to computing at RPI.

Among other things, people were interested in interactive (timesharing) use of the mainframe, instead of the batch system of OS/MVT. RPI was also running a timesharing system called "Alpha", but that was not a robust system, and it had trouble supporting even a dozen concurrent users. So, RPI went searching for a timesharing-based operating system that could stand up to the demands expected from RPI's environment. Operating systems that worked on IBM mainframes would have been preferred, but in fact we looked at operating systems from Honeywell and other vendors.

At one meeting of SHARE (a conference for IBM users), Wilson Dillaway had come across a group of guys talking about the MTS operating system. While that group didn't seem to be trying to sell anyone else on their work, Wilson was impressed with the discussions that were going on. Thus, MTS was one of the operating systems that RPI evaluated in early 1976. The committee unanimously recommended MTS over the alternatives, and MTS was running at RPI and available to users on a limited basis by fall of 1976.

(This quick overview is what I (Garance) remember of this early history. I was a freshman at the time, and I was one of two student representatives on that committee which selected MTS. I think that committee was chaired by Don Porter, but it may have been Wilson).


An e-mail from Wilson Dillaway

The following is from an e-mail exchange that Wilson Dillaway and Jeff Ogden had in August and September 2010:

On 9/29/2010 9:04 PM, Jeff Ogden wrote:
Do you mind if I use some of the information from your reply
below on the MTS archive site? -Jeff

On 9/29/2010 9:33 PM Wilson Dillaway wrote:
Sure!  I can't say that it's accurate; only a personal off-the-cuff recollection after many decades.  But if you can phrase it in those terms, then it works for me!  Wilson

On Aug 20, 2010, at 7:12 PM, Wilson Dillaway wrote the following in response to a question from Jeff Ogden.  The quoted text that Wilson is commenting on is Garance's posting that appears above.

Jeff,
Good to hear from you! All these memories! Maybe the best way
to respond is for me to intersperse below.
Wilson

From: Jeff Ogden
Sent: Friday, August 20, 2010 1:06 PM
To: Dillaway, Wilson
Subject: Can you confirm an old story?

Hi Wilson,
I'm still working on the MTS Wikipedia article (in my
retirement). Can you confirm this story? It sounds about right
from what I remember from back then, but it was a long time ago.
   -Jeff

MTS comes to RPI
http://mtswiki.westwood-tech.com/mtswiki-index.php/MTS%20at%20RPI

While RPI had been running IBM mainframes, student access was
mostly via batch jobs (which is to say "punched cards").

>>>> Yes, when I came to RPI in 1973, it was a punched card
shop, running MVT, with card readers and line printers
“accessible by students”. The main thing that students used
was the University of Waterloo’s WATFOR complier. <<<<

As one might expect, catering to students in engineering
and the sciences meant there was a large demand for computing
on campus. In 1975 RPI had purchased a used(!!) IBM 360/67
from Rice University, in an attempt to do an inexpensive upgrade
from their previous mainframe.

>>>> I was hired with the understanding that the 360/50 was to
be replaced with an IBM 370/158, and I was hired as the new
manager of the systems group to execute that transition
(coming from an insurance company in San Francisco which had
two IBM 370/165 systems both running TSO for 300 Cobol
programmers, in the process of upgrading to 168s). Having
arrived, I found out that the president had nixed the 370/158
deal due to university financial problems. We then went
looking for a used 360/65, and found a 360/67 at Rutgers in
New Jersey (not Rice in Texas), which we bought used. A 67
could emulate a 65, so nothing was lost, but it had this extra
capability, which was intriguing to me personally. I
subsequently persuaded RPI to run CP-67 (the predecessor of
VM/370) and multiple copies of OS/MVT on top of that, to make
the batch processing more efficient (VMware many decades
before its time!). I was the lead person working with the
CP/67 software, as well as being manager, so I came to
understand in much detail just how the 67 hardware worked
(UMMPS-level programming), which set the stage for
understanding what else might be possible. <<<<

That machine was simply not large enough to meet the demand,
and the students made it clear that they wanted more significant
improvements made to computing at RPI.

>>> The 360/67 wasn’t all that much slower than a 370/158 would
have been, which simply proves that the choice of the 370/158
as the upgrade path, originally, was based on what RPI thought
it could afford (before I came), and what IBM recommended,
rather than any judgment of what capacity was needed. <<<<

Among other things, people were interested in interactive
(timesharing) use of the mainframe, instead of the batch system
of OS/MVT. RPI was also running a timesharing system called
"Alpha", but that was not a robust system, and it had trouble
supporting even a dozen concurrent users.

>>>> IBM’s recommended time sharing system would have been TSO
on top of OS/MVT, which was nowhere near what IBM’s TSS was
designed to be. On the 360/50, Alpha supported terminals and
submitted batch jobs whose printed output could be inspected
online, but it didn’t scale, as Garance notes, and we never
offered it to students, only staff and a couple of faculty. It
was already in place when I arrived. I don’t recall where
Alpha came from; another university, I think. <<<<

So, RPI went searching for a timesharing-based operating system
that could stand up to the demands expected from RPI's environment.
Operating systems that worked on IBM mainframes would have been
preferred, but in fact we looked at operating systems from Honeywell
and other vendors.

At one meeting of SHARE (a conference for IBM users), Wilson
Dillaway had come across a group of guys talking about the MTS
operating system. While that group didn't seem to be trying to sell
anyone else on their work, Wilson was impressed with the discussions
that were going on. Thus, MTS was one of the operating systems
that RPI evaluated in early 1976.

>>>> I remember being the lead advocate for MTS back home, but
I had forgotten where I had first learned about it. I was a
frequent SHARE attendee, so Garance may be right. <<<<

The committee unanimously recommended MTS over the
alternatives, and MTS was running at RPI and available to users
on a limited basis by fall of 1976.

>>>> I wrote a paper comparing 17 different possible time
sharing systems, and recommended MTS as the preferred outcome.
That document is of course lost, and never was digital, I
suspect. As a committee, we made a trip to Ann Arbor, and were
graciously hosted by Mike [Alexander], as I recall. <<<<

(This quick overview is what I (Garance) remember of this early
history. I was a freshman at the time, and I was one of two
student representatives on that committee which selected MTS.
I think that committee was chaired by Don Porter, but it may
have been Wilson).

>>>>   Don Porter worked for me (hired after I arrived); I imagine
that we were both on the committee.  We hired him away from
Albany Medical College, were he had helped run a DEC-10?   
Or maybe I have my generations of DEC hardware mixed up!   
Anyway, he had time sharing experience on arrival.    Garance
describes himself here as a freshman, but he must have subsequently
worked for us as a student employee, and upon graduation we hired
him as a full time employee, and I gather he’s still there [yes, as of
October 2010 he is].  I had forgotten that he was on the committee,
though.

>>>> The process of bringing MTS online, testing it under CP-67 (which no other school may yet have done at the time??), and running MTS in test in parallel with MVT in separate virtual machines seems monumentally bold and risky now, but then we didn’t know any better (the theme of MTS overall, perhaps).    But that was the hardest part – getting it working under CP-67, gradually introducing it to faculty, and gradually migrating the load across, all the while continuing to run MVT batch in production, was with hindsight quite miraculous.    I think we may have overlapped (MTS and MVT simultaneously on that feeble hardware) for more than a year, but I’m not sure. . . .

>>>> One of the peculiar characteristics of RPI geographically is that Troy is just up the river from Poughkeepsie (or Armonk).    I vaguely recall that IBM hired a tremendous percentage of our graduates (being an engineering school), and we had IBM on the Board of Trustees.    They were not pleased with the choice of MTS, and when we later upgraded from the 67 to an IBM 3030 (??), we were under tremendous pressure to drop MTS and go with VM/CMS “so that our graduates would have the right skill set when they were hired”.    We resisted that, but even the students were hostile (they would go on summer internships to Poughkeepsie and come back singing the praises of VM/CMS and asking us to convert).   I left in 1984, and it was still a foaming issue as I left.

>>>> My feeble mind has few details, and many of them came back to me only in writing this.    So if anyone remembers otherwise, they are probably right!!  Much fun… Wilson



Comments