Limited Blocking Resource Sharing for Global Multiprocessor Scheduling

In this paper we focus on resource sharing in global fixed-priority preemptive multiprocessor scheduling. We consider the scheduling of a set of sporadic constrained-deadline tasks sharing a set of resources in a non-nested way. We propose the Limited Blocking - PCP, a novel resource sharing policy for global multiprocessor systems which aims to eliminate the negative effects of global multiprocessor PIP. We also develop the response-time based schedulability analysis for the proposed resource sharing policy. Finally, we analyze the efficiency of our work, compared with existing state-of-the-art resource sharing protocols.

I. Introduction

In the real-time systems realm, problems related to power consumption and limited further speed increase of uniprocessors have led to aroused interest towards the use of multicore processors for such systems. Moreover, the real-time research community put considerable efforts for developing multiprocessor scheduling algorithms assuming real-time applications consist only of independent tasks. However, typical applications include tasks that share resources like shared data structures which require appropriate techniques for ensuring mutual exclusion, such that only a single task holds the resource at all times.