Back to Home Page
Control Data Corporation - Experienceof Ed Thelen - September 1967 to 1972
Table of Contents:
- Leaving IBM
- Introduction to Special Systems Division
- Grumman Aircraft Contract
- Volkswagen Contract
- Wrap - Two Successful Projects, Why?
- Leaving CDC
- Short Stories
- Local Epilog
As I mentioned in Leaving IBM, the wife was unhappy about the house and lack of friends in Yorktown Heights (everyone figured we were like the other transient programmers - IBM = I've Been Moved ;-)) - and I had an open invitation to work for Control Data Corporation ( CDC ) in Minneapolis, an easy drive to wife's family and old friends. Also I had grown up 25 miles east, and was familiar with the area and had friends around. :-))
With the proceeds of selling our house in Phoenix, we made a down payment on a house on the west side of Pike Lake, New Brighton, Minnesota - about 3 miles from the new factory making the new Control Data 6600. ( Well, OK the folks to the south had a dock on the lake. Our lot didn't quite touch water, but we had a great view from the living room and basement ;-))
Introduction to Special Systems Division
The role of the Special Systems Division was to aid Control Data Corporation ( CDC) sales into special situations - such as adapting the hardware and/or software to unusual applications such as hybrid computing.
As I arrived, a one month CDC 6600 hardware class was starting for programmers. This class was taught by a red headed woman, of uncertain temper. We were taught how to optimize code for the unusual computer architecture. I was amazed by:
Our instructor really knew the programmer's model of the machine, and efficiently filled our heads. She was a good programmer, but she wanted to be a manager. - However no one wanted to work for her as she had an erratic disposition.
- User programs are protected from each other by use of base address and length registers controlled by the operating system.
- The speed of the logic, a 60 bit integer add takes 100 nanoseconds, a normalized floating point multiply in 1 microsecond.
- There are different "Functional Units" to do different instructions - in parallel - discussed here. Note that there were two MPY functional units, if one was busy, the other was used. A "Score Board" unit kept track of times and destinations.
- The machine does not have a halt instruction - a Peripheral Processor can "Exchange Jump" the the main machine to shift it to a different program. The Exchange Jump stored the machine registers into an area of memory, then loaded the machine registers from another memory area.
- You cannot single step ( single instruction ) the machine. Debugging the hardware must have been "INTERESTING" !!
- The main machine is totally controlled by the 10 peripheral processors (PPs) .
- One processor by convention controls the main machine,
- One PP to control the twin display CRTs and keyboard. The CRT displays were vector graphics so for instance you want to form an "A". You blanked (turned off) the beam and moved it to the lower left of the A, turned on the beam and moved it to the top of the A, and then to the the lower right, then turned off the beam, moved it to the bar, turned on the beam, and made the bar. - and so to paint the next next character. One PP was fast enough to paint all the required characters on both tubes, 60 times a second, and also monitor the keyboard.
- the other 8 PPs do I/O - responding to requests in a user program's address 0
The above is a brief introduction to the CDC 6600 hardware
- Software is needed to make that hardware do useful work.
The above is an introduction to the CDC 6x00 line - for "standard" customers.
- Burroughs, General Electric with the 6x5 line running GECOS, and the Control data 6x00 line running Scope, were all well ahead of IBM in the ability to:
- Call in needed components of the operating system as needed from disc rather have all the operating system in memory all the time. ( A smaller memory works just fine. )
- Run multiple jobs at the same time. ( Standard "Scope" could run seven user jobs, and the peripheral "job" at the same time - with the operator able to monitor and interact with each job - say mount a requested tape. )
- "Spool" input and output to/from fast disc rather than wait for slow peripherals. ( For instance, two or three card readers could be reading job streams onto the disc for later execution as jobs. At the same time, three or four printers could be printing output from previously run jobs. -
- IBM would have to use multiple 1401s to do the same support to the main processor, using tapes.
- IBM later used "HASP", a connected processor to do the same thing.
- Years later, IBM finally could do spooling from a main processor effectively.
- The design was optimized for minimum expensive memory. The software that directed the whole operation was in one PP, consisting of 4 K of 12 bit words. All of expensive main memory was available as User Memory.
Another 4 K PP could be directed by the boss PP to do print spooling. Memory was so tight that the print spooler PP needed to load an overlay from disc to do the BIG PRINT as the header for a print job. Loading the overlay took maybe 250 milliseconds - and saved expensive core memory.
Folks do not appreciate Seymour's money saving efficiency - they look at the bottom line, not counting up the efficiency of the machine.
- Speaking of memory - when IBM FINALLY got VMS working, the VMS operating code took up more memory than the largest CDC 6600 sold - If you ( IBM ) wanted to run user programs, more memory was required.
- The CDC 6x00 series was a scientific FORTRAN munching machine, it seemed tailored for that particular language. Eventually silly CDC corporate wanted to run COBOL on the 6x00 for business customers - such as the United Bank of Switzerland - and had a lot of trouble - needing a character oriented machine to be efficient.
- The original FORTRAN for the CDC 6x00 was called "RUN", but it had minimal optimization. Internal benchmarks tended to resembleDO 10 I = 1,10000 D = A * B + C 10 CONTINUEand the RUN compiler would make the machine do the loop 10000 times.
Other firms bought/made optimizing FORTRAN compilers which recognized that doing the inside of the loop once was enough - and made CDC's superior hardware look bad in benchmarks.
CDC had a crash project to add optional optimizing features to the "RUN" FORTRAN.
What was the role of the "Special Systems" organization?
To aid sales of the CDC 6x00 line to unusual, "Special" customers, usually requiring special interfaces to equipment such as
The above examples also had tight input, compute, output timing requirements.
- Analog computers ( the resulting called "Hybrid Computers" )
- Automotive test stands - such as for Volkswagen
- Aircraft flight test equipment - such as for Grumman Aircraft
We offered more than "best effort" offered by other companies,
We offered "Time Critical", a guaranteed input, compute, output, response time - for multiple data streams and multiple jobs -
So we also offered the software to fit the various requirements to the machine capabilities,
- and detect when
And we wrote special streaming logging to the system disc.
- the data stream requirements of a proposed job could not be fit into the existing data streams, causing the proposed job to be deferred until sufficient resources became available ( likely by other jobs terminating )
- the customer written code exceeded the running time interval stated in the job set-up. The customer written code was terminated - protecting the computing resource for the other users.
Grumman Aircraft Contract
Soon about ten people were formed into an evaluation team to visit Grumman Aircraft in Bethpage, Long Island, NY. Several of the people were from a Control Data subsidiary in Anaheim, CA - specializing in collecting data in CDC 1700 "mini-computers". (Grumman was paying for this "study"!)
We visited with Grumman engineers who were planning how to test a Navy jet fighter that Grumman had just one the development contract for. I believe it was for the E-14, now nicknamed the F-14 Tom Cat The request was for a three stream telemetry system to receive data from a prototype aircraft under test.
Interesting considerations were:
This was to replace a telemetry room full of chart recorders and human eyeballs used in the past.
- The major goal was don't crash the prototype aircraft. Great loss of time and money. I imagine test pilots were supposed to take care of themselves.
- But you want to test many situations (envelopes) as fast as possible.
Run a test, analyze on-line, and if advisable, go on to the next ( more vigorous ? ) test in the same flight.
- Support an on-line display driven by Grumman written FORTRAN programs which could display text and graphs, with keyboard interaction.
- All telemetry was to be via IRIG standards.
- NO data compression - even though a lot of the data seemed boring, if there is a flutter threatening a wing to come off - you want ALL the data points.
- To support on-line editing and compilation of FORTRAN programs, written by Grumman, (to analyze and graph the the data). (This eliminated IBM as a competitor.)
- Handle a new Laser Ranging Theodolite to determine the exact locations and times, and therefore velocity and accelerations of the test aircraft. And incorporate this data into the data streams.
- Retain the Time Tagging the data for analysis and display.
- All of the above data was to placed in real-time ( no missing data ) onto hard disk for quick retrieval, display, and analysis.
We sketched out a system suitable for the above requirements while at Grumman, and one of the group priced the hardware and software requirements. After a day of presentation, discussion, the negotiators went off - and hit a severe road block ( price of course ;-)). We tried to call the president of Control Data ( Bill Norris ) about what to do. We were told he was on a fishing trip and unavailable. So we packed up and headed for the airport and Minneapolis.
When we got the airport, our manager was being paged. Apparently the CEO of Grumman got through to Bill Norris ( CDC's CEO ), and resolved what ever. We were instructed to go back to Grumman at Bethpage.
Seems the big wheels have their methods of contact !!
Back at Bethpage, we got to serious work on the technical contract - data volumes, types, data rates, ... Very interesting - each page was initialed by each member of both sides, reproduced and distributed to all hands. We produced a document about 5/8 inches thick - the equipment and software carefully specified - but this document did not contain pricing ;-)) ( We working stiffs never did see the prices. )
The specified system included:
- 5 CDC 1700s to input data from the Grumman supplied telemetry equipment, convert/linearize the telemetry analog and counts to Engineering Units, and transmit the data to the main computer.
- One CDC 6500 (dual CDC 6400) with disk, mag tapes, printer, card reader, Calcomp plotters, ...
- Special disk driver to write data into large circular buffers in a time-critical fashion. And an access method to retrieve this data for application programs.
- Extended Core Storage (ECS) for rapid swapping of data and programs.
- A driver for the vector graphics terminal to allow graphic data from Grumman written FORTRAN programs to be displayed.
The above was the good news -
Then the ? inevitable ? politics started :-(( This turned out to be a rather unpleasant project. The manager doled out information only when squeezed -
Fortunately, at that time I received an offer to help bid the Volkswagen proposal, from the man who warned me of the potential bad review. I accepted that offer !! I didn't get to help install Grumman.
- The bid manager (near retirement age) did not get to manage the project.
He was replaced by a much younger man full of "success oriented schedule" and similar "pointy haired manager" ( ala Dilbert ;-)) slogans.
- The proposed software manager and his second were on another project - and did not want to get distracted by this contract.
- After the software manager and his second finally became available - they began trash talking the contract. I was in the hallway with the second when the Division Vice President came by. The second told the VP that the software part would have to be redesigned -
- The software manager and his second renamed all the software parts,
then we implemented the contract exactly as originally specified.
- I think to punish/embarrass me, I was assigned to write most of the acceptance test procedures and much of the acceptance test code. Little did they know this was a happy task for me - I got to keep in touch with large parts of the project ;-)) (Work with all the little secretive groups.)
- About the time the system was to be delivered to Grumman, I found out that the new software manager had written a bad letter of my performance.
(I verified this by getting into the personnel manager's desk.) ( Well, OK, I liked to know who was getting how much money, and did find out at all the places I worked at except IBM, usually just by opening/copying "available" notebooks.)
- I was irritated, and went on strike - talked my way to CDC 7600 hardware school for a week. My manager called, asking what was going on - we settled our differences a little.
The new project (and most of Special Systems) was being moved to Sunnyvale, California. We sold the house by the lake by owner, as usual. Put an ad in the paper, a co-worker topped all the bids.
But I kept track of installation and acceptance test problems at Grumman - definitely unofficially -
One member of the bid team for the Grumman project was from the CDC-1700 group in Anaheim, California. (We were using five 1700 systems as telemetry front end processors in the Grumman system. The 1700s decommutated the data and converted the data to "engineering units" for the CDC-6500 to process.) During acceptance testing that group had real troubles :-((
I was glad that I was in sunny California with a much happier group !!
- The CDC-1700 group had said that the Preston Analog-to-Digital convertors (ADCs) were biding had linearity to "3 bits" which he interpreted to be "3 bit counts" - a little over one bit error plus-or-minus from "perfect".
The Preston convertors as delivered were plus or minus "3 bits" or about plus-or-minus seven counts from "perfect". They were being really being pinched by Grumman for this deficiency. Grumman, being in the cut-throat aerospace business, was a really tough customer - squeeze you on everything.
- The CDC-1700 computers were having strange reliability problems. They finally made "smoo diagrams" which plotted how the computers were working on a two axis plot.
- one axis ambient temperature (min through max specified)
- the other axis the voltage of the power supplied to the 1700s, min-max
In happy theory the computers should work at all points in the graph area -
- but there was an area in the middle of the graphs for two 1700s where they failed !! :-((
It took that group over a month to get that resolved - all very expensive !!
Feb 2015 - I just got reminded of an interesting story
One of the Grumman guys gave me a copy of Edward White's body telemetry as White did the space walk -
on an 8 inch wide, 6 pen, strip chart - with notations -
- made about 2 years earlier
He explained the various pen recordings:
- respiration (chest circumference)
- EKG traces (heart rate)
- blood pressure
- and a couple more I forgot
and he explained where White got into the following:
- started breathing fast (breathing too fast, too much, can make you dizzy)
- heart started racing
- where White reported getting dizzy, and requested to come in.
Apparently "walking" in no gravity,
- floating, disoriented - where is up? -, and what ever
takes some getting used to -
I don't know if they made that big "swimming pool" (excuse me Neutral Buoyancy Laboratory) to acclimatize astronauts was before or after Edward White's adventure..
-Ed (them was the dayze) Thelen
PS :-(( I lost that strip chart within a year. Maybe showed it off too much :-((
PPS :-(( This was the same Edward White killed in this fire :-((
So - of the 150 Control Data Special Systems Division families which were offered a move to sunny ( and somewhat crazy ) California, about 100 took the offered free trip to look the place over. (Fifty were so interested in ice fishing or what ever that they didn't even take the offered free trip !! )
Of the 100 families who took the trip, about 70 moved with Special Systems to the San Jose area.
By about three years later, about half of the 70 families who had moved to sunny California had returned mosquito infested, iced over Minnesota. Maybe it is the Minnesotans who are a little crazy !!!
In any case, we all had house price shock !!! In Minnesota, for $20,000 we had a nice house on a suburban lake. (And we could borrow a neighbor's snow blower.) In California, for twice the price we could have a slightly larger tract house.
So - shortly after moving in, about 10 of us were off to Wolfsburg, Germany to talk with Volkswagen. (You understand, the proposed customer was paying for this "study".) Along on the trip were:
- John Sansom, who had warned me (above) and offered me a job on the Volkswagen team.
John also brains behind the analog simulation language MIMIC.
- John Munson, an analog computer guy adapting to the digital world, slated to be the Volkswagen project manager if we got the order.
- Gary Peterson slated to be the software supervisor, if we got the order.
- Some hardware guys
We met Helmut Mach, the account salesman.
The only other person I have known that seemed to have that same quality of deep calm force was Helmut Mach, the Control Data salesman in Germany that developed this major computer order from Volkswagen. The word about Helmut was that he had been captured by the Soviets in the battle of Stalingrad, survived the horrible German/Soviet conditions, and as a captive of the Soviets until 10 years after the war had varnished/polished furniture in factories for the upper Communist Party members. Of the 250 thousand Germans captured at Stalingrad, about 13,000 (one in twenty) survived to eventually return to Germany. Helmut had that same calm, forward looking, no nonsense, inevitable pressure.
Interesting person - I noticed when our initial meeting was about over, he put his hands into his pants pockets. As we left we each got a WARM handshake -
Basically, we bid a scaled down Grumman system (above) to the German manager who had made a name for himself getting a wind tunnel to test VW "Bug" automobiles. This was another interesting guy - He said that it was inevitable that extra expenses would develop as the project proceeded - but that could not exceed 10% of the contract price - his "Iron Reserve", to be disturbed only in high stress.
(By the time the contract was delivered, very little of his "Iron Reserve" had been touched - we bid the project very well :-))
We got the project, and again I got to write the acceptance test specs and much of the acceptance test code.
This was a happy project. Six engineers came from Wolfsburg, to learn the system and help assure we were doing what was needed. They were a lot of fun.One of the VW engineers was Matt ?Zurcha? who had been on the ?Austrian? cross country ski team. My wife and I were involved with the Ski Touring Section, in the Sierra Club. Matt could ski circles around us all. One winter trip, the club and Matt went ski touring near Mt. Bishop in the east Sierras. We drove as high as we could, but could not leave the cars there. Matt was one of the volunteers to drive the cars down to official parking and "catch up". Soon Matt "caught up" as we were struggling up in deep snow - waving happily as he passed by. The rest of us finally got to some frozen lake to set up snow camp when we heard shrieks of joy and yelling from high above. Matt blew into the snow camp maybe a half hour later, he had gotten way up on Mt. Bishop, and was celebrating on the way down. "The rest of us" were properly humbled. He was actually a really nice guy - short, scrawny, a good enough mathematician, but could he ski !!!!
As part of the contract, VW sent 6 "test stands" to Sunnyvale to assure that we could talk to them properly. (In this business, surprises are mostly unpleasant.) There was a MAJOR hang up in U.S. Customs. They Wouldn't be convinced that these test stands would be returning to Germany with the system. After some political/diplomatic mumbo-jumbo ( $?? ) they were FINALLY released from the port of entry.
So, after a year or so, the system was ready, and shipped. This time I "helped" install it in Wolfsburg. By the time we arrived, the hardware was up and running - and the software/performace testing was very smooth. I wanted to go home, but my German tester counterpart wanted to go to the Munich Olympics, and I was pressured to stay handy.
The wife came over to Germany and we went to Switzerland to see the mountains. We were at a restaurant when suddenly people stopped talking, and the radio was turned up.
What was happening ??? ??? After a while, someone explained that some Muslims had killed many Israeli athletes at the Olympic Village.
Wrap - Two Successful Projects, Why?
Several times I have watched How Can Software Be So Hard? by Professor Martyn Thomas CBE, and many other software project presentations, and admired that we were more successful than most - and wondered why.
Not only that, but I have met many people who have worked on major software projects, and hated the experience. I had (mostly) a good time, well paid for having fun :-)
Some considerations seem to be:
- I came in to a group that was competent and had created and shipped much of the special (time critical) software before. Individual special projects were based on a known, proven base.
- The top manager (managed several project managers) was experienced in business, with some hardware & some software experience.
- He was not pressured, nor did he pressure, to get every possible contract. We could walk away if the project did not fit our experience, our cost/time estimates, or the pricing too sharp.
- He encouraged customers to send knowledgeable people (typically half a dozen with some software experience) to help with the effort, and be able make modifications after system acceptance. We did not want to be in the software maintenance business.
- These people could provide practical guidance about things that were not fully specified in the carefully crafted contracts we developed.
- (I presume they reported back to their HQ good progress, and helped keep their HQ out of our hair.)
- On the Volkswagen project (above), the customer sent six types of test stands for us to verify that our hardware and software worked as expected before everything shipped, California - to Germany. There was even a door slamming test stand, as well as engine test stands. One engine test stand was programmed for an engine run from (The customer was responsible for dealing with the customs officials of each country - non-trivial.)
While I was working on the Volkswagen project (above) several things happened.
So, having a family to support, and liking a nice income stream, I started looking around. Times were good - I was young, frisky and knowledgeable/experienced/skilled :-))
- One of Special System's core businesses, hybrid computing, was declining. The price performance of digital computing was getting VERY competitive with analog computing
- MIMIC (above) and other analog simulation languages were being used instead of painful set up of analog computers
- Attempts to have digital computers do high speed set up of analog computers with fancy switches was marginal - many intermittent electromechanical problems
- Control Data, Special Systems, had made a major contract with Swiss banking, using the CDC 6xxx family, (wonderful FORTRAN floating point munchers) were painfully slow and awkward at COBOL, and this project was not doing well. Control Data had no concept of Data Processing, JCL - Job Control Language, checkpoints, journal tapes for recovery, banking, business data processing, ...
Control Data was at a standing start in a world dominated by IBM who had been thriving in that business for decades !!!!
- No nice big fat juicy projects seemed available for bidding, execution. I had made a number of trips to prospects with John Samson - nothing promising -
- I became convinced that Control Data did not know how to make money on the software tools we were using. Too much tweaking, not enough standard stuff.
- And most frightening of all, Control Data pulled all their sales force out of Special Systems Sunnyvale back to Minneapolis !!! It was hard enough to educate the sales force as to our special benefits - Including "Time Critical" guaranteed performance, rather than best effort "Real Time" computing.
RCA (a force in computers at the time) was hiring into their Educational Division in Palo Alto. On the way to the interview, in the RCA building, I bumped into my old nemesis, Tagtow from the Grumman project - he said he was a manager there - I don't even go the few feet further to the interview.
I was lucky !! A few months later, RCA shocked the computer "community" by announcing the closure of all their Computer Operations, including the Educational Division.
(Unlike General Electric they didn't even sell the computer operations off, RCA just closed the doors.)
I also applied at General Electric Atomic Division in San Jose. They wanted someone who knew the GE 645 and its GIOC (General Input/Output Controller). Since I was involved with the diagnostic tests for the GIOC, I figured I was one of the most qualified candidates in the world for that job. I talked with some HR person, who knew nothing useful at all. We looked cross-eyed at each other - he said I was unsuitable. I asked if I could talk with Dr. Weinberg, who had placed the ad. The HR guy said "NO". I later called Dr. Weinberg, who said I had to go through HR.
Again I was lucky !!! The atomic energy business in the U.S. collapsed with in a year !!!
I then got into Measurex and was happy for about 18 years :-))
In my son Carl's words, I was no longer an "iterant programmer" ;-))
First Experience with Jet Lag
When we were studying/bidding the Volkswagen system, Gary Peterson, the proposed software supervisor, arrived at Wolfsburg a few days late. He hurried from the plane directly to a customer meeting. The poor fellow was so confusiled that we had to lead him from the conference room to a safer place. :-|
After a day or two he could and did make useful contributions.
What Year is it? from my eldest son
"Most glaringly, you didn’t start with CDC until summer of 1967, so your subtitle needs a slight modification. It was the summer of ’67 that we moved from New York and lived on the farm. I remember very distinctly that I had just turned 7 and just finished first grade when we packed up, put the cat in the car (always put a leash on the cat when letting him out for a potty break, it once took a *long* time to get him back into the car) and drove to the farm in Wisconsin. A GEMCO store had just burned down, with large loss of life due to lack of “panic hardware” on the doors – which opened inward, not outward – and for months the twisted scorched steel framework was visible from the freeway. We were in MN for two years (my 2nd and 3rd grades) and moved to California in the summer of 1969 (with a side trip to the World Expo in Spokane Washington, where we used our car to help a police officer block off a side street for the fireworks display and were rewarded by being able to be almost right under the display as fireworks debris rained down around us – what fun).
"BTW: a few years ago I dropped by the Pike Lake house and found that all the yards are now fenced in. The view of the lake is different."
My reaction - Sad, just to keep the snow mobile folks from roaring through the back yards.
And from youngest son
"> It was the summer of ’67 that we moved from New York and lived on the farm."
"Wow, that must have been quite a year. I was born in January of '67 in Phoenix, AZ. Dad, couldn't you stay still for 3 months? It's called 'settling down,' not 'driving your family from coast to coast to coast.' :-) "
Guilty as charged :-((
In retrospect, sounds inconsiderate - Also Carl was born just after I finished GE-225 school in Phoenix, Arizona and had moved to Somersworth, New Hampshire. I was helping install the computer in Somersworth while Carl was being born in the local hospital. :-((
Various family members have kindly refreshed my never good memory for dates. :-))
"Chuck Urmson" wrote:
> CDC is dead and buried.
Indeed - you reminded me to add an epilog to my CDC story -
A word picture
A few years after leaving CDC as a lost cause - and happily at Measurex - I was curious about how local CDC was doing - I went to the old Special Systems building - no one there knew of CDC Searching in the phone book White Pages, remember them? , I found a phone # and location of CDC in the San Jose area. At a small shopping center office, the front door propped open, It was fall, and the fallen leaves were blowing in the door. Inside the receptionist said this was the CDC office, but didn't know what CDC did. She said that if I waited a little while a salesman would be back and tell me -- I asked about the leaves blowing about - She said the janitor would remove them in the evening - - -
If you have comments or suggestions, Send e-mail to Ed Thelen
Back to Home Page
Web page started Nov 2011
updated through Feb 2018