Arrival of the Net at UofM (Draft)

By Thomas C. Jacobson

In the beginning the UofM did not want to be connected to what was then called the “ARPANET”… although there were offers, and several faculty wanted it, no administrator could see the value of being connected to a national academic network. They always asked, “What would you use it for anyway?”

By about 1982/3 Joe Thomas and I had configured two LSI-11s to communicate between the main campus and the Lauderdale facility using software from Dave Mills at the University of Maryland called “Fuzzball”. This was the first TCP/IP link on campus and ran at the amazing speed of 1.8Kbps for the demonstration between a lab in Fowell Hall and the Lauderdale computer facility. The idea was to use the Fuzzballs as routers connecting serial links, Ethernet, and NSC Hyperchannel to the CDC Cyber mainframes at the Lauderdale facility using TCP/IP software for the CDC systems developed at Tektronix. At the same time I successfully campaigned for the purchase of a SUN-1 workstation by walking around campus showing off a borrowed Sun CPU board and claiming that the 68000 processor was equal to a speeding DEC VAX 750! By early 1983 we received SUN-1 serial number 128. With the arrival of the Cray-1 I promoted the idea of getting a Sun workstation to communicate with our newly arrived Cray-1 at the UofM with Sun founder Bill Joy during a USENIX meeting in Paris France later that year. (Bill eventually promoted the idea at Cray Inc. where work was later done to interface a Sun directly to NSC Hyperchannel and to implement gateway software for the proprietary Cray Operating System (COS).)

By the winter of 1984/5 Research Equipment Inc. the new holding company for the Cray-1 and the soon to be delivered Cray-2, and owned by the UofM, was selected by NSF to provide supercomputer services to national researchers. To provide access, REI submitted a proposal to NSF for connection to the ARPANET at 56Kbps. The last of the antiquated DEC Unibus BBN 1822 link controllers and Unix 4.2BSD emulation software was purchased at great expense and installed on the VMS VAX 780 then serving as a frontend to the Cray 1, thus establishing a connection to a BBN IMP (ARPANET hub) at the University of Wisconsin. At that time access to the ARPANET was only from that VMS VAX, although some limited fixed routing was used for testing.

With the pending the arrival of the Cray-2, the pressure was on to implement a campus network so that UofM researchers could access this $20M+ resource. Together with Tom Walsh of the Supercomputer Institute, I demonstrated to a skeptical community that a TCP/IP network was possible using early Bridge Communications TCP/IP terminal concentrators, TCP/IP software for the PC from MIT, and Excelan TCP/IP for the VAX VMS Cray-1 frontend system was viable, although there was great pressure to use antiquated punched card Remote Job Entry (RJE) technology. I also proposed and demonstrated that a Codenoll fiber optic Ethernet backbone could connect campus buildings together.

I also convinced Eric Benhamou, then VP of Engineering at Bridge Communications, to implement IP routing on their equipment, and several of these Bridge IP routers were purchased to link the campus to the Lauderdale facility at 56Kbps. The Supercomputer Institute also purchased several SGI workstations with Excelan TCP/IP Ethernet controllers in them, and we shipped one to Bridge for testing with the prototype IP router. The other SGI workstations were shipped to the UofM where they were used for the NSF funded summer program during the summer of 1985, and then distributed out to the principal Institute of Technology departments. After the NSF summer program was over, Tom Walsh hired Randy Smith to act as network administrator, and deployment of the optical network began. Since UofM plant services refused to install the fibers in a timely manner owing to construction work and union problems, a team of “high energy” physicists and I broke into the steam tunnels using the best Richard Feynman lock-smith techniques, and installed the fibers during several Sundays that August, climbing up and down 100 ft. steam shafts in stifling heat. Fortunately no one was killed.

During 1985/86 period I was lucky enough to convince Scott Bertilson, Stuart Levy, and Joe Thomas to join the REI (later called the Minnesota Supercomputer Center, MSC) networking group that I was privileged to lead.

By the fall of 1986 REI and the Cray-2 supercomputers moved to a new building on Washington Ave, the ARPANET connection was backhauled there, and connected to a VAX 750 running 4.2BSD Unix that had both an Ethernet controller and a Hyperchannel connection to the new Cray-2. Later the ARPANET (now called NSFNet) connection was migrated to a Proteon router and the link protocol migrated from the weirdo BBN 1822 link protocol to standard HDLC, and the VAX 750 then only served as a router between local Ethernets and the Cray-2. Sometime after that the VAX system(s) were replaced by Sun servers, also connected to Hyperchannel to act as routers to the Cray-2, which, as a Unix system, had native TCP/IP support.

At times, there were in excess of 80 simultaneous Telnet sessions on the Cray-2 system(s), originating both on the UofM campus, and nationally. There were many experimental network software developments, such as applications using Remote Procedure Call (RPC), a specialized version of the Unix editor Vi to reduce character-at-a-time overhead on the Cray-2, and network file services.

The MSC networking group’s legacy includes significant contributions to the development of the Internet we know today. It touched on everything from the nuts-and-bolts to founding of the Minnesota Regional Network (the first Internet Service Provider in Minnesota) and generally fertilizing the Internet in Minnesota. It made contributions to the development of many Internet technologies, including collaboration with individuals at the UofM Microcomputer group who eventually developed Gopher.