This is the next-to-last (the fancy word is "penultimate") column of this computer history series. The last one will cover my last years at IBM, which weren't very grand. My time in the Supercomputing area described herein was my last chance for some really good work.
IBM traditionally was not into the supercomputing race. Supercomputers are very custom, huge, somewhat cranky computers that are used mostly for large-scale scientific use, simulations, and modeling. There wasn't a market for too many of them at that time, the development costs were high, and the staff needed to support them disproportionately large compared to normal mainframe computers. For years, IBM ceded that market to Cray computers and other smaller companies.
But IBM had a visionary in Dr. David Wehrly. David was an MIT PhD., very bright, iconoclastic, and a born leader. When my time in large mainframe computers was grinding down, I heard of David's shop from some friends. David and I hit it off, and I was able to arm-twist my way out of my current job and into his area.
David's area at that time was very much of a "skunk-works" - a term used to describe a small and loose organization of very bright people working on leading edge ideas. And the people were bright - I am pleased to say that I helped that by bringing in more people who fit that characteristic. We had a large lab to play in, lots of computer "toys" and a lot of latitude in our jobs. I remember taking an afternoon off to see the movie "Tron" that had a lot of early computer graphics in it - something you would NEVER do in mainstream IBM.
Of course, we had some serious projects going. As IBM was not ready yet to build its own supercomputers, we worked at integrating IBM mainframes with Cray computers - using IBM computers for I/O (input/output) and the Cray for pure computation. We worked at the National Center for Atmospheric Research (NCAR) in Boulder CO, and we had some really great times there. We were also working on Parallel Fortran - the ability to use the multiple processors in a typical IBM mainframe working together on a single problem. Most of this work was done at Cornell University.
We finally were able to get our own supercomputer project going. A single computer processor can only go so fast - but if you could harness the power of dozens or hundreds of processors working together, you could obtain huge speed increases. Of course, not all computing problems can take advantage of parallelism, as it was called, but many could. Besides having a very compact processor "box" structure, we also had a proprietary high-speed link between them, which is the key to using all that performance. We were also leaders in using UNIX-like systems at IBM.
Unfortunately, Dr. Wehrly did his job too well, in one sense. Our work, and parallelism in general, got the attention of the large mainframe development organization. They decided that what was needed was a true-blue IBM approach, rather than a skunk-works, and within a fairly short period of time our project was terminated in favor of a similar but more mainstream supercomputer. The area as we know it was dismantled, and Dr. Wehrly was edged into early retirement.
That was also the time that the mainframe market collapsed. IBM closed our plant, the huge facility at Kingston, NY. I remember talking to many of my friends from the skunk-works about leaving IBM and forming our own company, but we couldn't get enough people in the boat. So, I worked in the new area for a brief time to transition things over, and then took off for the final segment of my IBM career in Raleigh NC.
(Bob Seidel is a local computer consultant in the Southport - Oak Island area. You can visit his Website at www.bobseidel.com or e-mail questions or column ideas to him at bsc@bobseidel.com. For specific inquiries, please call Bob Seidel Consulting, LLC at 278-1007.)