Novelty in the Nemesis Operating System

Abstract

The opportunity to develop an operating system from scratch, even in a research environment, does not arise very often. One needs the inclination, a motivating proposition, and the resources to carry the development through. However, the results can be very rewarding; starting with a blank sheet of paper encourages novelty and allows conventional wisdom to be challenged. In the early 1990’s the Systems Research Group at the Cambridge Computer Laboratory began a project to develop an operating system to support the processing of continuous media. The team wanted to investigate the provision of guarantees of predictable performance for a dynamic mix of applications generating, playing, and/or processing audio and video information. The work was carried out within two serial EU funded projects. Providing predictable performance to applications entails giving them the resources they need at the time they need them. If one concentrates simply on the processor(s) as a resource, then it is tempting to think of this problem simply as a scheduling problem. While scheduling is key, it is not the only issue. It was recognised that uncontrolled resource interference between applications—denoted as resource crosstalk—could arise as a result of the structure of the operating system over which the applications ran. An operating system, computer scientists are told from a very early point in their immersion in the subject, is a program (or set of programs) that controls the resources of a computer system, protects users from one another and provides services above the mere hardware of the system. They are also told about hard