Towards Dynamically Adaptive Operating Systems

Abstract

The highly dynamic nature of distributed processing in a network of computers is a major factor determining the overall system performance and therefore must be taken into consideration when the operating system algorithms for distributed resource management are designed. Such algorithms must be able to detect situations of dissatisfactory resource consumption and adjust the system behaviour in order to improve its performance. In this paper we propose a general framework for integrating such algorithms into an operating system which is then able to dynamically adapt itself to changing environments. The approach is based on the cooperative effort of three major entities, an observation base for collecting information, a knowledge base for detecting areas of improvement and an execution base for taking appropriate actions. The feasibility of our solution is demonstrated by presenting adaptive algorithms for distributed load sharing and replica assignment.