Multics > Sites
12 Mar 2006

Site History: UC & ACTC

Location

University of Calgary, Calgary, Alberta, Canada

The university had a department called Academic Computing Services. They set up the Advanced Computing Technology Centre in 1983. In 1988, a spinout called ACTC Technologies was created, which got its own Multics machine and provided services to Honeywell, Bull, and other Multics customers. ACTC was later known as Perigon, and was later bought by the CGI Group, Calgary. (This is a different ACTC from the Alberta Carpenters Training Centre or the Alberta Cattle Title Committee.)

First Installed

UC: 1978
ACTC: 1983

Configuration

UC: started out with two L68 CPUs, 3MW of memory, and 12 MSU451 disk drives. It eventually grew to 5 DPS8/70M.

ACTC: initially had 1 CPU on same floor as UC. The system grew when they were doing Multics support, and then shrank. In 2000, ACTC had 1 CPU / 1 4MW SCU / 1 IMU / 7.4GB disk.

The Sale

[Bob Petrowski] I met with Dr. George in Ontario before he joined the U of Calgary and took him to HIS Phoenix, introducing him to the Honeywell board directly. This was a major turning for HIS when Dr. George took over the U of Calgary activities.

[Bob Petrowski] I worked with the VP of finance Harvey Bliss, tuning the final contracts and paperwork to make it happen. I helped push the old CDC system on its last Friday (still dead) out the back ramp and assisted in pulling cables and flooring to get that first MULTICS system up and running for Monday operations. I kept the food and drinks coming.

[Bob Petrowski] I followed up and held weekly meetings with U of Calgary staff to get rid of rumors, learn and hear about problems and convey them to tech support to bring about a uniform understanding and clean up early settling in problems.

[Bob Petrowski] I visited every lab and researcher on campus growing in their education and understanding of how to use the MULTICS systems. From Engineering to Biology to Astronomy to Physics.

[Bob Petrowski] I helped grow the center adding memory, software applications, additional CPU's, and over 1,000 CRT's for campus wide connectivity to the growing user base of MULTICS. I later joined the ACTC migration and selling of services with Dr. Ron George.

[Bob Petrowski] There are many stories, about our people, from out techs who made sure the hardware ran, to our users who were encouraging us to push things further, by developing products. Please add an honorable mention for Tony Hwang who worked many extra hours on de-bug, he was not alone, there were others.

Site Analysts

[Norm Barnecut] Don Mengel was the first SiteSA. He must have been around for a year and a half staying at a hotel at Honeywell's expense. They must have forgotten about him.

Jerry Kelly (79-80), Michael Armstrong, Alf Burnham, Ricarda McDonald

FE: Tony Hwang.

UC Folks

Director: Ron George

Director of Computer Services: Collin Marx

Associate Director, Advanced Computing Technology Centre: Brian Wescott

System administrators: Norm Barnecut, Steve Harris, Allen Haggett, Hal Hoover, Tom Oke

Support: Marc Donovan, Dan Freedman, Kevin Jameson, Ned Kittlitz, Gary MacIsaac, Ward Rosin

Users: Deanne Dearing, James Gosling, Dave Schroth, Russell Schulz, Greg Woods.

ACTC Folks

President and CEO: Ron George

Director: Arun Gatha

Senior Vice President of Research and Development: Brian Wescott

Staff: Ward Anderson, Henry Blackmore, Richard Bubric, Nazrat Durand, Dean Elhard, Mike Flegel, Dan Fudge, Jim Gee, Rick Gray, Carol Hinatsu, Harold Hoover, Douglas Howe, Suzanna Huen, Alfred Hussein, Mazen Itani, Kevin Jameson, Kirsten Kallstrom, Dave Leskiw, Michael Mabey, Maureen Mallmes, Mary McGuire, Tom Oke, Doug Robinson, Cameron Roe, Charles Rohs, Ward Rosin,Dave Schroth, Phuong Vu, Rob Waters, Betty Wong, Ron Wright, Louis Zimmerman, Doug Zwick

Notable developments

[Norm Barnecut] The university had a department called Academic Computing Services. ACTC was spun off from that. I think it was when ACTC became a separate company that it became known as ACTC Technologies. University Technologies Inc. is a technology transfer company wholly owned by the University of Calgary.

ACTC was contracted by Bull to take over Multics maintenance in January 1989. One reason for this was that there were tax advantages to having some development done in Canada.

Application areas

Computer science.

Anecdotes

The benchmark, in 1978, specified running some large number of users doing exactly the same thing, perhaps a BASIC compile and run, with specified response time. Multics shone especially well under such conditions: all the users shared a single copy of all the procedures in memory. After the sale, when the university began adding users, they got nowhere near the same number before performance fell off drastically, partly because they were all doing different things, which increased memory contention and paging.

See Tom Oke's paper presented to HLSUA Forum XXXI October 12, 1980 about tuning Multics.

The University was very unhappy with Multics performance when the system was installed. Honeywell was forced to send Multics tuning experts to Calgary multiple times and provide extra hardware. There were also serious reliability problems post sale, which led to several visits from Honeywell engineering. Some of the problems were due to the computer room environment.

[Kevin Jameson] My understanding of the sale/post-sale problems are as follows. Multics was the first timeshare machine on campus, replacing a CDC Cyber 172 Kronos timesharing machine. Benchmarks were done on the Multics machines to match machine capacity with computing needs on campus (user job loads, scientific number crunching, etc.). Multics was installed to meet the benchmarks.

[Kevin Jameson] But no one understood that Multics was a (1) secure (2) timeshare machine, and what that meant in a university context. The whole campus (10000 people, 5000 users?) was always logging in and logging off. Since the CPU login quantum for each new login was something like 1 or 2 dedicated seconds of CPU time, the whole machine (5 cpus) was constantly thrashing with the login/logout load. So no real computing could be done.

[Kevin Jameson] It didn't help that the communication front end (DCS Distributed Computing System) was terrible, terrible. The DCS collected terminal input from users, and then routed characters into Multics through two Arpanet interfaces. The DCS could support about 250 users max, I think, in the early days. Then it would crash.

[Kevin Jameson] To give you a flavor of working on Multics in those days, if you were sending a piece of email or typing in a document, you would save your file at the end of each line of text, because there was a high probability that the DCS or Multics would crash before you could enter two lines without saving. Even in a lightweight editor like "ed/vi/qedx" or the like, it often took five minutes to enter a line of text and save it. (This was under heavy load in the daytime hours, of course.)

[Kevin Jameson] The machines would work fine until around 830-900 am, when the load would spike up with new logins, and then computing would be painful for the rest of the day.

[Kevin Jameson] In the end, the director of computing was fired over the fiasco. An article was written in Business Quarterly describing the whole mess. Ron George was the new replacement.

[Norm Barnecut] Did I ever tell you about storming up and down the hall yelling THVV? It had something to do with bugs.... I see one to three character variables. They are i, ii, iii, j, jj, etc. It must have been in the answering service.

Final Shutdown

UC: 1993
CGI: July 2000

Information from Kevin Jameson, David Schroth, Norm Barnecut, and Bob Petrowski.