ELS: An Hard Real-time Scheduler for Homogeneous Multi-core Platforms

Abstract

This paper presents a scheduler, called ELS (based on the combination of EDF, LLF, and SJF algorithms), for hard real-time tasks in a homogeneous multi-core platform with the aim of increasing system performance (success rate). To obtain a suitable scheduling, ELS exploits the combination of three EDF (Earliest Deadline First), modified LLF (Least Laxity First) and SJF (Shortest Job First) scheduling algorithms in different task conditions. The core principle behind the ELS is choosing tasks based on the laxity ratio to the worst case execution time. The ELS scheduler was simulated in SimSO standard simulator in two cases, i.e., quad and eight-core platforms. The results of the experiments on 100 randomized tasks generated by the RandFixedSum algorithm showed that the success rate has improved on average 10%, 14.1%, 10.8%, and 33.3% compared to the EDF, SJF, LLF, and RM (Rate monotonic) respectively in a quad-core platform. Also, the success rate in an eight-core platform has improved about 31.6%, 21.6%, 6.6%, and 38.3% compared to EDF, SJF, LLF, and RM respectively.