Work Packet Manager: exploiting heterogeneity in cluster computing

In expensive high performance computer systems combined with high speed networks and machine-independent communication libraries have made cluster computing a viable option for parallel application programs. Heterogeneity is quite common in this environment, especially in processor architectures: rapid technological change means that as elements are added to or replaced in such a system. In this paper, we analyze a technique for exploiting this heterogeneity to improve the performance of parallel applications in cluster system. The Work Packet Manager (WPM) library is a general and extensible work scheduler, which mimics Cilk's dynamic scheduler, for MPI. The experimental results demonstrate the performance benefit of the WPM library in heterogeneous cluster environment.