Optimized Scheduling for Parallel Computing Environment

Abstract

The increasingly parallel, heterogeneous, and powerful computing infrastructure demands efficient scheduling of tasks over computing nodes to fully utilize these computing resources. For an optimal scheduling strategy, it is important to consider all the parameters that have an impact on the overall execution time of a problem having tasks that may run in parallel.This work proposes an optimal scheduling algorithm for a class of problems that can be represented by a directed acyclic graph. The proposed algorithm is static in its nature and considers heterogeneity of processing elements. A comparison with existing algorithms under different scenarios also shows improvement in overall execution time.