Scheduling Parallel Virtual Machine in Xen Based on Credit

Abstract

With the development of multi-processor system, parallel program becomes a very important direction of software development. But in virtual machine systems, the performance of concurrent program scheduling is poor. In order to improve the performance of scheduling parallel applications in Virtual Machine (VM), this paper raises a parallel schedule algorism called PSA, which is based on Credit scheduler in Xen VMM. This new algorism support for scheduling the parallel Virtual CPUs (vCPUs) synchronously, it reduces the effect of asynchrony of parallel vCPUs by the interrupt of I/O bounded workloads, and it keeps the synchronization of parallel vCPUs when asynchrony emerges in the process of scheduling. The evaluation shows remarkable improvement of the PSA in scheduling the parallel vCPUs with different workloads, compared to the original Credit scheduler, co-scheduler and UVF: the demand-based coordinated scheduler.