Multics > Sites
26 Jun 2024

Site History: AFDSC

Location

United States Air Force Data Services Center, The Pentagon, Arlington VA, USA.

First Installed

November 1973

Configuration

This site had 6 systems: four were multilevel secure, two were unclassified.

[MAJ Jennifer Lyon] At the time I left the Pentagon in 1988, we had 6 systems:

4 were multi-level (D, H, M, & Z)

2 were unclassified (T & X). T & X could be run as a single system or split to be 2 separate systems.

System Z was brought operational soon after I got there in 1983. Sys X came along sometime around 1986.

Some data from 1986

Application areas

[CDT] Big Consistent System users. Lots of database-type work. Financial chartsy-graphsy applications, budget analysis, program analysis and evaluation. Air combat simulations and their graphical interpretation. Lots of "critical reports for the Secreta ry."

Honeywell Marketing

Jesse McDuffy, Bill Cratty, Jane Brady

Site Analysts

Field Engineers

AFDSC Personnel

[John Ata] The following AF personnel worked on Multics and were fairly knowledgeable about the OS (more so than some Site SAs):

[Ed Rice] Joan Shields was the head of the entire system support shop at AFDSC.

Umar Khan said he was a "Master Multician" in the Pentagon from 1982 to 1990 in a 1993 USENET posting.

[Eric Swenson] I looked at this page, and noticed that absent was any mention of quite a few of the AF personnel who maintained the 6 Multics systems we ran. Some of the people I remember are:

We all worked there when Major (Dutch) Blake was in charge of systems maintenance and system administration.

System Administrators

Final Shutdown

Applications

The Wartime Manpower Programming System (WARMAPS) was designed and implemented by General Research Corporation starting in 1978. It is described in a final report and in Wartime Manpower Planning System (WARMAPS), ADP System Users Manual, DOD 1100.19-M.

[Jim Densmore] I was on the TAC BRAWLER team at AFSA. We didn't just run graphical interpretation programs. We ran the simulation programs themselves on Multics as well. The over 200,000 line FORTRAN program that ran at that time was eventually ported to many machines using ANSI standard FORTRAN 77, and is over 500,000 lines now. Because of the object-oriented technology we had in play at that time (though we certainly didn't call it that), the program is still in use and still quite maintainable.

Anecdotes

[CDT] I once had a GSA security guard storm into the terminal room with his gun pointed at me when the "BEL" beeper on a Tektronix 4014 got stuck on. Does that count?

[CDT] The operators were relatively untrained. One noticed disk errors from a certain drive two days in a row in the early afternoon. He logged them for the FE without bothering to go into the next room to check out the drive. When the FE arrived to run tests on the drive, he noticed fiberglass dust all over the room from the ceiling repair work -- except for two clean shoe-shaped areas on top of this particular drive. The contractor was standing on it and drilling into the ceiling.

[CDT] The ceiling work was constant, due to the incredibly fragile chilled water system. In 1973, the system blew, damaging nearly every computer and device in the room. After that, the Air Force bought plastic covers for the equipment. Next time the system blew, they put the covers on, but didn't shut the equipment down. Of course, it burned out. From then on, they simply strung up dropcloths like huge tents whenever the water blew. It was like working in an electronic orchard infested with gypsy moths.

[CDT] AFDSC was one of the pioneer sites for Halon fire control. They needed a lot of tanks to flood the huge room, and located them centrally in the same room as a D270 disk farm. They girdered them up at head height to one of the walls. Then they foolishly rigged up the sensors so that any ONE going off would trigger the system. A couple months later, one sensor falsed. The force of the discharge blew the tanks, girders, and all right off the wall and across the room, completely scalping the D270's.

[CDT] GSA personnel upgraded the systems from core bulk store to MOS bulk store by wheeling the MOS box into the Top Secret area, cabling it up, and wheeling out the core bulk. The core bulk had been sitting in the hallway for three days before someone discovered that it had never officially been wiped of classified data.

[CDT] In about 1975, there was a huge five-alarm fire after hours that took out a corner of the Pentagon. It was blamed on spontaneous combustion in the cafeterias, which were being renovated. I got caught in the smoke leaving the building, and had to leave by an unfamiliar exit. It wasn't until I got home and saw the fire live on the news that I realized that the computer room was under that particular corner, and there had been no fire bells and no warning. I called the chief operator and informed them there had been a fire going on over his head for about two hours. He said he had only found out about 15 minutes ago, "when all the water started coming in." I told him to put the plastic covers on and shut everything down. He replied that there were plastic covers for everything BUT the Multics equipment. It seemed to be a good time to take some vacation, so I did.

[CDT] The one about the feral cat that lived under the raised flooring is not particularly Multicious. All the other good stories I know are classified.

Field Engineering Stories

[Robert Welsh] I worked for Honeywell 1973-1975 at AFDSC, mostly on the GE-635 systems and the EIS 6000, but occasionally on the 6180 and of course on its peripherals.

Regarding [CDT] GSA personnel upgraded the systems from core bulk store to MOS bulk store ...

I was one of the many Honeywell FE's who worked around the clock through the weekend to swap out the core memory for the new MOS memory. We ran 3 shifts per day, 7 days a week anyway due to the WWMCS contract. We had a lot of trouble because we were not as diligent as perhaps we should have been with anti-static procedures. We were also delayed because connecting and disconnecting the 3 phase 240VAC had to be done by GSA, we were not allowed to mess with it at all. Anyway, the job was completed sometime late Sunday evening/early Monday morning as I recall. I was a midnight shift guy, but I would often hang around in the mornings to hob-nob with the daytime personnel, plus I wanted to see how the new memory would be greeted by the customer (the AF). Major Bartos is the man I remember asking the question about the core memory now in the hall: "Did you zero that memory before disconnecting it ?" "Oh yes sir !" was the reply (not made by me, I was a by-stander). And then, the deadly question: "Were there any Air Force personnel there to witness it ?" Of course not, so we had to re-install the core memory cabinets and zero them with AFDSC witnesses.

Regarding [CDT] In about 1975, there was a huge five-alarm cafeteria fire after hours ...

I was working 2nd shift that evening, heading back to the "printer farm", when I saw an Air Force personnel RUNNING towards where the systems' consoles were. I had never seen anyone run in that computer room, so I followed to see what was happening. Water, dirty, sooty water was gushing down from the ceiling in several places, through soggy ceiling tiles that were now starting to fall, and AFDSC operators were scrambling with tarps placing them over the disk drives first and the system cabinets second (the system cabinets were very tall, and were not easy to cover). I went straight to the circuit breakers that controlled each string of 8 disk drives. It was a panic for sure! I got my manager out of bed and he came to the site right away. We spent the next several days replacing crashed heads on the disk drives, hundreds of heads. My understanding at the time was that we pretty much used up the entire stock of spare disk heads on the Eastern seaboard that week, and emergency orders had been placed to CDC (MPI), the manufacturers of the drives. I worked double shifts, replacing the heads, aligning them and returning units to service, only to have contaminated packs be introduced which once again crashed the heads. Because it was a multi-day emergency, I started to cut some corners. Proper procedure was to replace the heads, test by loading a scratch pack, then loading the alignment pack (which cost nearly the price of a new car back then) and aligning the heads. After a few drives, I was confident and skipped flying the test pack, going straight to the alignment pack. After doing a few of these, I felt uneasy I had been pushing my luck, so I flew scratch pack and bzzzzzt! I had installed a head wrong (A-up into a B-down slot?). Lucky for me it was a scratch pack (of which we had plenty) and not the alignment pack. They might have buried me under the floor for the cat to eat.

Air Force makes long-term commitment to Multics

Article in PRINTOUT, Honeywell Large Information Systems Division, Vol. 12 No. 40 September 30, 1982.

The United States Air Force has announced a $57.1 million award to Honeywell Information Systems, for life-cycle funding of the Multics systems in the Air Force Data Services Center in the Pentagon. Multics was chosen initially because of its ability to handle data with multiple security-classification levels simultaneously and safely. The Air Force already has 12 Multics processors installed at this site, and recently ordered four 8/70M processors. They are expected to order additional 8/70M equipment in 1983, making this the world's largest Multics installation. In addition to funding these equipment purchases, the Air Force award will pay for Honeywell hardware maintenance and services, software training and system analysts to be stationed on-site.

DATAMATION Article

Volume 33, No 7, January 15, 1987.

U.S. TO REPLACE MULTICS UNITS

WASHINGTON, D. C. -- The Pentagon this quarter will announce a request for proposals to replace 21 Honeywell large-scale computer systems, of which 15 run Honeywell's Multics operating system. How the Pentagon will replace the super-secure Multics is still up in the air, but Major Chuck Bowen, who's in charge of the bid, says, "Honeywell isn't going to support the equipment after 1988." Honeywell group vp Eugene Manno hasn't heard about the bid, but says work is progressing on HVS 6 Plus, a new OS that will take some Multics functionality and is expected in 1989. Unlike Multics, which runs on a special version of the 36-bit DPS 8, HVS 6 Plus will run on the 32-bit DPS 6 Plus.

Most of this info is from Chris Tavares (CDT), with updates from Jennifer Lyon, Ed Rice, Jim Densmore, Eric Swenson, and Robert Welsh.