Scheduling algorithms for cache effect optimisationin partitioned systems

Abstract

Partitioned systems have emerged as a result of the increasing computer processing power and application complexity and the needs to achieve more secure systems. The use of cache memories into realtime systems has been consolidated in order to mitigate the gap between CPU speed and memory performance. This paper introduces an extended task model in order to model the effect of the cache memories in the computation time of the tasks of a partitioned system. We also propose and evaluate four different modifications to the EDF policy that are intended to improve the CPU utilisation of real-time systems taking into account the additional knowledge provided by the extended task model.