A Novel Efficient Index Model and Modified Chord Protocol for Decentralized Service Repositories

Abstract

Distributed service repository systems are now being deployed widely rather than the traditional centralized service repository systems, issues such as information redundancy and service failures are still prevailing in existing models and search algorithms. Inverted index structure is a widely adopted model for service management and retrieval in service repositories, but the inverted index shows considerable redundancies which can significantly increase the retrieval time in large- scale distributed repositories. Chord is a widely used protocol for service retrieval in structured P2P networks. However, it is not sufficiently abundant for efficient service discovery and retrieval among all the nodes, since each node's finger table only contains the clockwise direction key-node information. To this end, this paper develops a multilevel index model for service maintenance and retrieval for decentralized service repositories, for the purpose of reducing the service retrieval time. In addition, a modified algorithm, called MLBChord, is deployed in this paper, for improving the search efficiency. Experiments conducted on a realistic simulation environment shows that our developed model reduces the retrieval time more than 50% to that Chord with inverted index, thus achieves an effective search retrieval time.