Optimized Scheduling of Periodic Hard Real-Time Multicore Systems

Abstract

Multicore systems were developed to provide a substantial performance increase over monocore systems. But shared hardware resources are a problem as they add unpredictable delays that affect the schedulability of multicore hard real-time systems. In recent years much effort has been put into modelling interference and proposing scheduling techniques to mitigate its negative effect. Using one of these models, we propose a scheduling technique, based on Integer Linear Programming (ILP) that, in combination with a task to core allocator, not only achieves a feasible schedule but also reduces the interference produced by shared hardware resources in the context of hard real-time multicore systems. The evaluation shows how both the interference is reduced (≈ 83.47%) and the schedulability is increased (≈ 12.25%) compared to traditional heuristics.INDEX TERMS integer linear programming, optimization, partitioned systems, real-time systems, static scheduling I. INTRODUCTION

Schedulability Analysis in Fixed-Priority Real-Time Multicore Systems with Contention

“…The allocation problem is recognized to be NP-hard in the strictest sense [5]. In multicore systems, the way tasks are allocated to cores largely determines the global interference [6]. Some of the main allocation techniques are heuristics based on bin packing, such as First Fit, Worst Fit [7,8], etc.…”

“…Regarding scheduling strategies to reduce interference, the recent work [6] proposes the Rolling Horizon MILP Algorithm (RHMA), which not only achieves a feasible schedule but also minimizes the interference generated by shared hardware resources within the context of hard real-time multicore systems.…”