Logo image
針對週期性即時工作的公平品質率排程演算法
Thesis

針對週期性即時工作的公平品質率排程演算法

陳柏元
Masters, 國立清華大學, 資訊工程學系
2003

Abstract

即時系統 回饋機制排程演算法 平衡回饋系統 週期性即時排程演算法 real-time systems reward-based scheduling balanced-reward systems periodic real-time scheduling
Reward-based scheduling provides graceful degradation for real-time systems such as multimedia applications and iterative-refinement numerical algorithms. A reward-based task is composed of a mandatory part and an optional part which only executes after completion of the mandatory part. Traditional reward-based scheduling algorithms address on maximizing the total reward in a system. Such an algorithm may result in an unbalanced system where some tasks receive results of superior quality while other tasks only receive minimal acceptable results. In this paper, we present an optimal balanced-reward algorithm such that each task receives the same quality ratio. We first discuss the case where each reward function is strictly increasing and invertible. A mathematical approach is presented to determine the optimal quality ratio. We next develop an efficient generic solution for general cases where some reward functions are not invertible. We conducted experiments to compare our algorithm with a set of other reward-based algorithms to demonstrate its effectiveness. The experimental results show that our algorithm effectively and efficiently develops an optimal balanced-reward schedule.

Metrics

1 Record Views

Details

Logo image