An efficient service search strategy in decentralized e-business environments

Search for services is one of the basic operations in decentralized e-business environments, and in such applications, reducing search latency is no longer the only motivation. Other important issues should also be considered, such as load balance, selection fairness and quality evaluation of data items. Based on the general peer representation and search methods of random walk in unstructured P2P topology, this paper introduces a novel mechanism of cached multi-dimensional virtual cluster of service peers, and then proposes a strategy of data replicating and a 2-phase searching approach. Experiments based on the prototype indicate that multiple target values, such as smaller search size, better search quality, load balance and selection fairness can be obtained comparing with general strategies of data allocation and search.