Adaptive Runtime Management of Spatial and Temporal Heterogeneity for Dynamic Grid Applications

This paper addresses the runtime management of spatial and temporal heterogeneity in both, scientific applications and geographically distributed resources in Grid computing environments. The targeted applications are large-scale dynamic Grid applications which require large amount of computational resources typically spanning multiple sites and exhibit very long execution times. An adaptive runtime management framework with a hybrid space-time runtime management strategy (HRMS) is proposed by combining adaptive application partitioning and resource scheduling techniques. HRMS defines a set of flexible mechanisms and policies to adapt to the state of both applications and resources. As a proof-ofconcept, a simulator for key features of this framework is being developed. Preliminary evaluation demonstrates that HRMS scheme improves performance and provides better speedup while using fewer resources on average.