Multisite job scheduling models in computational grid environment

Cooperation of massive parallel processor systems in different management-domains lies on the design of multi-domain job scheduling models. Three multisite grid job scheduling models, named local, sharing, multisite job scheduling models, are presented to overcome the heterogeneity and low bandwidth in computational grid. The job performance models on different scheduling models are analyzed in theory. The analysis results show that when the sites are homogeneous, multisite job scheduling should prefer the local and single site resources to multisite; when the sites are heterogeneous, these heuristics are invalid so that new heuristics should be designed to achieve the sub-optimum solutions.