Load-balancing strategies in distributed workflow management system

摘要

The grid technology has been used to solve complicated scientific problems for a long time. Since the heterogeneity of gird, we need an effective method to dispatch the tasks to computational units with different process abilities. In this paper, we first propose a new measurement standard for workload, and then three load-balancing strategies, namely RANDOM, shortest processing time first (SPTF) and longest processing time first (LPTF), are compared in order to find out a relatively optional one. The LPTF rule shows better performance when the number of workflow instances increases beyond a specific threshold value.