An agent-based grid computing infrastructure

The conventional computing grid has developed a service oriented computing architecture with a super-local, two commit scheduling strategy. This architecture is limited in modeling open systems with highly dynamic and autonomous computing resources due to its server-based computing model. The use of super-local scheduling strategy also limits the utilization of the computing resources. In this paper, we propose a multi-agent based grid computing infrastructure to tackle the above issues, while provide reasonable compatibility and interoperability with the conventional grid systems and clients. Compared with the existing grids, the new infrastructure is leveraged by the intelligent agents, and therefore is more efficient and flexible for open systems.