Scheduling multi-component applications in heterogeneous wide-area networks

In this paper we present a scalable scheduling heuristic for several common classes of multi-component applications (meta-applications). We consider this scheduling problem in a wide-area heterogeneous computing environment, or metasystem. The heterogeneity and scale of the computing environment and the heterogeneity of the application make this a challenging problem. We have studied the performance of the heuristic in simulation and the results are encouraging. Completion times for three common classes of meta-applications were within 10-20% of optimal on average with a worst-case variance of 60%. The results suggest that effective scheduling of meta-applications is possible, if sufficient application and system resource cost information is provided.