Multics Site History: USGS

Location

US Geological Survey, Denver, CO
US Geological Survey, Menlo Park, CA
US Geological Survey, Reston, VA

First Installed

Denver, 1976
Reston, 1977
Menlo Park, 1977

Configuration

bulk store

The Benchmark and the Sale

[Ron Riedesel] I noted the dearth of info about one of our biggest Multics sales, the one to USGS, and so decided to submit a few of my own meager recollections about it.

I believe that Don Heitt was the FSO sector manager, Elmer Clegg was the FSO sales manager, and the salesman who closed the order was Bob ?(who? 30-ish, tall, sandy hair, lanky, mustache).

I don't remember much else about USGS, other than the benchmark effort to win the business. On that score, I recall that Mike Grady and Bob Mullen came to Phoenix for an extended period of time to tune the system, and that Ed Rice (and maybe somebody else?) and a project manager named Carl Stanek came from FSO to work on it. The Camelback Road system was only available nights and weekends for much of the effort, so we spent countless overnights and any number of 48-hour sleepless stretches on weekends, over a number of months, making trial runs. Mullen was a Coca-Cola junkie and the stock value of the local bottling company skyrocketed during his stay. An overnight run meant a MINIMUM of 6-8 cans of Coke. A weekender was 2 cases. The local Coors distributor also made out pretty well.

The benchmark had very tricky timing and sequencing requirements, and the keys to making it work were: 1) a combination of new general-purpose tuning features added by Mullen to the scheduler, 2) a set of very specialized scheduler hacks Mullen installed on the Camelback system (that he said were too specific to the USGS scripts and timing requirements to be usable anywhere else), and 3) some FEP changes made by Grady that allowed stream-specific management of the (simulated) terminal workloads. I think we failed to pass the benchmark on the first couple of attempts, and made it on the final pass by the skin of our teeth.

If I remember this right, during the USGS live benchmark runs, we actually came in second in performance, but the other vendor that finished ahead of us (Sperry?) was caught cheating, driving at least one of the required line printers via a second mainframe, surreptitiously cabled in from another room.

Lucky we hadn't thought of it ourselves. We did just about everything else to win.

The Printout article announcing the sale thanks Allen Berglund, Don Mengel, Wade Myers, Ed Wallman, Bob Franklin, Charlie Clingen, Mike Grady, Bob Mullen, Ron Riedesel, Lacy Johnson, Rich Chouinard, Jerry Cahoon, and Harold Van Sant for support on the benchmark.

[Ed Rice] I dug deeply enough into my archives to find a copy of the USGS proposal submitted by HIS FSO, and it is dated January 26, 1976. That puts the start of the benchmark period back into about August of 1975, and Mullen's participation to around mid-October (I may be off by a few weeks). Working from that, I'd place the pxss mods that included the early elements of workgroup scheduling at right around Thanksgiving of 1975.

[Garry Kaiser] Other Marketing names: Steve Balacco, Bill Cooper, Dick Spacholtz, Tony Winstead.

USGS Folks

System administrators: Mary Ward (Reston), Kathy Schabacker (Reston), Dan McGlothin (Reston), Pedro Cadenas Jr. (Reston), Bob Terrazas (Denver), Wayne Schomburg (Denver), Tony Marshall (Menlo), Ken Pope (Menlo).

Operations: Ron Hopkins (Reston).

User Assistance: Jim Fisher (Reston), Sandy Grado (Menlo).

Users: Jim Mueller (Menlo), Pedro (Pete) Cadenas, Tom Hyde, Pat Doherty, Pat Murphy (Menlo), Noel Roulin, Jim Browning (Menlo/Denver), C. Andrew Shepp (Denver), Patrice Gunderson (Denver), Joseph Cain (Reston), Rex Sanders (Menlo), Bill McChesney.

Notable developments

[Paul Benjamin] Continuum (ancestor of forum) was written by USGS-Menlo user Pat Doherty. It was picked up and improved by Mike Auerbach and installed on System M to coordinate MRDS changes for the EOP bid response.

remote printer support

[Ed Rice] Very large MRDS database for National Mapping Center (USGS Topographic division).

Anecdotes

[Andrew Shepp] In the seventies, I worked in Denver with 2 other Multicians on a contract with USGS (Department of Interior); it was PL/I programming and access to the new relational database. I also discovered the Multics APL source code and played and learned very much. The Primary Contractor was Potomac Research Inc, almost immediately gobbled up by EDS.

[Andrew Shepp] This was supposed to be a "rock" project where we would load then access the results of rock core drillings conducted by USGS, but the project was managed by an active-duty, uniformed, US Air Force Bird Colonel, so the DOI might have been a cover. We were told there were only 4 Honeywell Multics machines in existence, I was amazed at the 128 bit pointers which could point to any page in virtual memory on any Multics CPU or storage or tape device. Once I got APL running, I made the usual entry []WA (Workspace Available), IBM's VS APL or APL under TSO would give a number like 32,000 (bytes). Multics gave me a floating point number!

[Andrew Shepp] I remember my username as ashepp.rtr and Miss Gunderson as PGunderson.rtr.

See Rex Sanders' USGS-M Recollections for more stories about the Menlo Park Multics.

[Dave Jordan] My recollection is that most of the major issues USGS had with Multics had to do with Fortan - either it didn't work right...or it wasn't fast enough. I recall spending a lot of time tracking down and reporting Fortran issues.

[Dave Jordan] I do recall one emergency trip to Menlo Park when their system wouldn't boot. I'm a bit hazy, but I think maybe they'd run out of online dump storage space or something? I flew out, spent about 5 minutes at the console, figured it out (whatever it was!) and headed back to Denver.

Final Shutdown

Menlo Park , 1983
Reston, 1984
Denver, 1985

Information from Ron Riedesel, Mary Ward, C. Andrew Shepp, and Dave Jordan.

Site History: USGS