Experiences with Mainframe Computers: A Personal Journey

The UNIVAC 1108: A Step Back in Time

My first exposure to a mainframe computer was through the UNIVAC 1108. This behemoth of technology had 262K 36-bit words of core memory, equivalent to approximately 1.125 million bytes. For a young graduate student in the 1970s, it was a revelation. It featured 16 accumulators and 15 index registers, which seemed like a luxury compared to the first-generation microprocessors like the 6800 and 6502, which had just one or two accumulators and one or two index registers each.

The 1108 was known for its speed, with single-precision instructions taking just 750 nanoseconds. This was a significant improvement over the 6502 microprocessor found in the Apple ][, which operated on 8-bit data and took about 5 times longer to execute a typical instruction. However, the 6502 was still vastly simpler in design and purpose.

Handling Data and Input on the Mainframe

Programming on the UNIVAC 1108 involved using punched cards. Jobs would be submitted in batches, and you would return to collect the results the next day. The cards contained not only your source code but also any necessary libraries, input data, and job control language that would guide the computer and its operator through the process. If there were any steps that required manual intervention, such as mounting a tape, these would be specified in the job control language as well.

A Later Encounter with the Mainframe: IBM System 370

Several years later, I had the opportunity to work with a mainframe computer again when I was contracted to write the backend for a Pascal cross-compiler targeting the 8086. The compiler ran on an IBM System 370, and I do not remember the specific model number.

A large part of my work was done remotely, dialing in over a 300-baud modem using an ADM-3A terminal. This setup was rudimentary by today's standards but was typical for the time. However, once a month, I would fly down to Phoenix for a week of in-person development, using a local IBM 3270 terminal with TSO (Time-Sharing Option) to edit and compile my compiler in real-time.

During the day, the compiling cycle could take several minutes, as the mainframe had been recently upgraded with 8 MB of RAM, but it was shared among dozens of online users. This often resulted in noticeable delays. Therefore, I developed a habit of coming to the system early in the morning when I was the only user on the system, and the full 8 MB of RAM was available.

This period of early morning work was highly productive. Compiles would typically take just 2 or 3 seconds at such times, allowing me to get more done before 7 am, when others would start logging on and the system began to slow down due to increasing user load.

Conclusion

My experiences with mainframe computers have been both educational and eye-opening. These systems, with their massive memory and complex architecture, set a benchmark for what was possible in the world of computing at the time. Working on early computing machines like the UNIVAC 1108 and the IBM System 370 has given me a deeper appreciation for the evolution of technology and the challenges faced by early computer scientists and engineers.