In January 1978, I started work for Sperry Rand (later Sperry, Sperry Univac, and Unisys), in the group that supported NASA’s Communication (NASCOM) organization at the Goddard Space Flight Center (GSFC) in Greenbelt, Maryland; our office was off-site a few miles away. My badge is shown below (with my Social Security Number obscured; we were so innocent in 1978).
NASCOM supported all of NASA’s manned and unmanned missions. When I started, NASCOM included three world-wide networks of voice, low-speed teletype circuits, and high-speed data circuits. The Sperry staff was divided into three groups for these different types of communications, and I was assigned to the 3760/CPx group supporting the high-speed network.
When I arrived, NASA was just preparing for the launch of a satellite (Seasat) that would use new high-speed communication capabilities, and everyone in the group was very busy supporting last-minute testing. With nothing much to do, I spent my time reading manuals and looking at code, trying to understand what these computers did, and how they did it.
The operational complement was two 3760s; two more served as “hot” backups (meaning they could be switched on-line immediately in the event of a problem with the primary set), and a fifth for development and test activities. A 3760 was the size of a desk. In fact, you could sit at it; all of the processor and memory circuits were in the space occupied by the drawers in a normal desk. There was a terminal and keyboard on the desktop, and a pair of tape-cassette drives for loading software. Next to the desk was a refrigerator-sized cabinet housing the specialized I/O processing, called an SIOC. Each cabinet had 20 or 30 circuit boards, about 18-24 inches square, each with two or four processors. Each processor connected to a single high-speed data line. In 1978, high-speed meant something up to 56,000 bits per second (56kbps), about the speed of the last dial-up modem you might have had. These data lines carried data in packets of two sizes: 1200 bits or 4800 bits. This was different from the low-speed data lines which handled variable-length messages containing individual characters. Although the I/O was handled by the SIOC, the memory of the 3760 held the data buffers that the data passed through. When a packet came into a buffer, the 3760 examined its header and determined which output lines it should be sent out on; most data packets were sent to multiple destinations (up to six, as I recall). The memory of the 3760s was quite limited, with only enough space for about 100 packets at any moment, supporting about 30 data lines.
This routing function is essentially the same as what your cable modem does. It has multiple connections (mine from Verizon has a main network connection to the cable, four ports to plug in ethernet cables, and a WiFi network), and can accept input data from any of the connections, and route it to the appropriate output connection. In this way, any of the computers, phones, tablets, and printers in my house can communicate with any of the others and the internet, without needing dedicated wires connecting each one to all of the others.
In addition to the 3760s, NASCOM used Univac 494 mainframe computers. These were large computers, each contained in several refrigerator-sized cabinets, containing powerful processors, memory, large tape and disk storage, and many I/O connections. There was one online and a hot backup that was also used for development of software. The 494 ran the development software for all of NASCOM’s computers.
Previous: Sperry/Unisys/GSFC – Overview