A DMR Fair Algorithm for Realtime Scheduler

Abstract

Earliest deadline first (EDF) is one of the most popular soft real time (SRT) scheduling algorithms being widely used. Slack time and overrun are the two key factors that will affect the performance. Reclaiming slack is a favorite approach that has been introduced to EDF. Adaptations of EDF tend to concentrate on reducing average deadline miss ratio (DMR) and improving average tardiness, while paying little attention to fairness of the service quality between tasks, which is important for streaming media applications. A DMR fair EDF algorithm is proposed in this paper, which use expectative execution time as a factor, in addition to its deadline, when a task¿s priority needs to be determined. In comparison with SLAD and BACKSLASH algorithm, the proposed algorithm brings a fair deadline miss ratio among tasks, with an acceptable ADMR and ATRD between SLAD¿s and BACKSLASH¿s.