Abstract

Multicluster Grid has emerged as a major approach for people to pursue substantial computational capability improvement to support large-scale compute-intensive applications. However, it is challenging on how to efficiently manage application execution across the participating clusters, since the participating clusters are geographically distributed, heterogeneous, and selfadministrative, and the network connection provided by the Internet is vulnerable in security and unpredictable in performance. The challenges include how to discover resource information, how to handle application workflow, how to allocate appropriate resources, how to make application execution reliable to endure system failures, etc. This paper proposes cluster abstraction to uniform describe and access resources in multicluster Grid. After investigating various cluster resource management systems, a reference description of cluster abstraction is presented to cover cluster computing characteristics, hiding individual cluster details for use. Following this concept, a deployment service is discussed. This deployment service is one of the essential components in ongoing resource management development for multicluster Grids. 1.

Citations