Trust Management in the P2P Grid

Abstract

Peer-to-peer (P2P) Grid Computing has gained its popularity in many large scale distributed applications. Within such an environment, trust management is a complex and difficult task since the Grid harnesses a large variety of geographically distributed resources in virtual organizations (VOs) across multiple domains. This paper proposes a reputation framework for trust management in the P2P Grid. In order to efficiently collect locally-generated feedbacks and aggregate them to yield the global reputation scores, the proposed model integrates two known concepts: distributed eigenvectors method and the central method. A novel FARW algorithm based on DHT network is designed to efficiently aggregate global reputations within one VO, and the ring algorithm is suggested to elect the coordinator, which is responsible for reputation calculation across multi-VOs. To validate our system, a context-related reputation function is designed to facilitate an efficient mechanism for service provider selection in this P2P Grid environment. Network simulation experiments show that significant performance gains can be gotten using this framework.