Challenges in the Implementation of MrsP

Abstract

The transition to multicore systems that has started to take place over the last few years, has revived the interest in the synchronization protocols for sharing logical resources. In fact, consolidated solutions for single processor systems are not immediately applicable to multiprocessor platforms and new paradigms and solutions have to be devised. The Multiprocessor resource sharing Protocol (MrsP) is a particularly elegant approach for partitioned systems, which allows sharing global logical resources among tasks assigned to distinct scheduling partitions. Notably, MrsP enjoys two desirable theoretical properties: optimality and compliance to well-known uniprocessor response time analysis. A coarse-grained experimental evaluation of the MrsP protocol on a general-purpose operating system has been already presented by its original authors. No clear evidence, however, has been provided to date as to its viability and effectiveness for industrial-size real-time operating systems. In this paper we bridge this gap, focusing on the challenges posed by the implementation of MrsP on top of two representative real-time operating systems, RTEMS and LITMUS$^{RT}$. In doing so, we provide a useful insight on implementation-specific issues and offer evidence that the protocol can be effectively implemented on top of standard real-time operating system support while incurring acceptable overhead.