Parallel Job Scheduling with Time-varying Constraints for Heterogeneous Multiple-Cluster Systems

摘要

Parallel job scheduling in multiple cluster system is a critical aspect to facilitate the processor cooperation. However, co-allocation, heterogeneity and time-varying emerge as tough challenges for the design of multiple cluster job scheduling models and algorithms. This paper presents a new multiple cluster job scheduling scheme based on the time-varying job performance model for a dynamic, heterogeneous multiple cluster environment. Four multiple cluster scheduling algorithms based on different heuristic resource selection strategies are introduced. Experiments indicate that the scheduler and the algorithms are effective and perform better than the single cluster and other multiple cluster scheduling algorithms.