PRIME: a modular architecture for terminal-oriented systems

The architecture of most interactive systems is based on the general strategy that suitable terminal service can be provided by a central processor that is timemultiplexed among all the active terminals. In order to achieve adequate response time in an interactive environment, the CPU is usually time sliced. Other major system facilities such as I/O channels and secondary storage units are also shared among the users, and multiprogramming techniques are employed to keep all the major system resources as fully utilized as possible. An operating system is usually developed which performs these functions as well as supervising the terminal communications, implementing a system-wide filing subsystem, handling user commands, etc. The result of combining these and other functions into a time-sharing operating system is a highly complex software system which transforms what is basically a batch processing computer structure into a multi-terminal system with significant limitations that are an outgrowth of this strategy. While a failure can occur in any section of the hardware or software, we know that hardware failures are more likely to occur in the electromechanical and core memory sectors than in solid state logic, and that software failures tend to be concentrated in the more complex areas of code. Failures of hardware components may require modification of the operating system in order to regain operational status since the allocation strategies may need more than parametric modification when system resources are affected.