Workstations and Supercomputers

While attending some classes at Stanford in 1977 I explored their computer science department in Margaret Jacks Hall, and visited the wonderful semi-circular building of the Stanford AI lab (SAIL) west of campus many times. I met a number of people such as Len Bosak, who later founded Cisco, etc. On later visits to the campus I became interested in their effort to build a sort of follow-on to the PARC Alto workstation, but based on the new 68000 processor, and using other technologies developed at Stanford, PARC, and SRI such as the mouse, bit-mapped displays, and Ethernet. Back at the UofM I proselytized what was being done at Stanford to anyone who would listen, and carried a Forward Technology SUN-1 board around explaining how it was almost as powerful as a DEC VAX 750, and why bit-map displays, small talk, etc. was the way of the future. I passed out copies of the early drafts of overview papers, which were later published as CSL reports like TR 229 shown here.

Later, I met with  Sun founders at a Usenix conference in Toronto, and after we got our Cray-1, I met with Bill Joy at a European Usenix conference in Paris. I described how tightly coupling workstations to supercomputers using the Berkeley Unix TCP/IP stack that Bill was responsible for might enable interactive fluid flow visualization, etc.

Eventually I got the University to purchase one of the first Sun-1 workstations (s/n 128, probably the 28th one made), which included a mux card (which happened to be made by a company founded by a cousin of mine!) and connected dial-up modems so people could explore the Sun Unix implementation. As I recall, the Sun-1 still had bugs, and I spent time on the phone with Andy Bechtolsheim debugging early hardware, I remember he sent me a number of EPROMs to fix things. Seeing our success, the UofM CSci department, and others, purchased Suns.

When I became Director of Networking for the Minnesota Supercomputer Center Inc. we obtained an NSF grant to implement a BBN 1822 link on a DEC VMS Vax to the ARPANet IMP at the UofWisc. in Madison. We built a campus Ethernet backbone, and routed through the VAX to many types of workstations. In time we used Sun systems with NSC Hyperchannel as front-ends to our Cray-1 and Cray-2 systems, and later as routers when the Cray-2 supported TCP/IP directly. In addition, we routinely supported in excess of 80 telnet remote terminal login sessions using Bridge CS/100 terminal servers to the Cray-2, not to mention FTP, and a special version of vi that reduced per-character interrupts to the Cray-2 by mapping local vi screen editor actions on the workstation into ed line editor commands (rvi).

The Supercomputer Institute was given money to purchase eight SGI workstations, and others on the UofM campus had purchased Sun, Apollo and HP workstations. We made a special effort to connect them all to our Cray computers with TCP/IP, and eventually Remote Procedure Call at the application layer. I had the idea of setting some of them up in front of the Cray-2 for this picture….