Agent Based Resource Discovery for Peak Request Periods in Peer-to-Peer Grid Infrastructures

Abstract

One of the fundamentally required services in the grid environment is resource discovery. The discovery involves the search for appropriate resources that match user requirements. An efficient mechanism for this service still remains a crucial problem especially within a dynamic and scalable environment such as the grid. Majority of the proposed solutions based on centralized and hierarchical approaches suffer from shortcomings ranging from single point of failure to network congestion. In this paper, we propose a resource discovery mechanism that relies on the activities of an agent during peak request hours in a peer-to-peer (P2P) based grid system. The agent searches and learns the paths to requested resources with associated maximum rewards. These paths are managed by the super-node for subsequent resource discovery requests. We evaluated the performance of the proposed approach against some resource discovery approaches. The results show an improved performance in the proposed algorithm over the Time To Live (TTL) and and Adjacency List and Ant Colony Algorithm (GAA).