pService: Towards Similarity Search on Peer-to-Peer Web Services Discovery

We present the design of pService system for P2P Web services discovery supporting similarity search with its locality-preserving feature based on modified Skip Graph. In pService, the most important property (e.g. service name) of a service is viewed as the indexing key of a peer node. The other properties are extracted as multiattributes which would be mapped into one-dimensional indices using HSFC (Hilbert Space Filling Curve), and assigned to membership vector of the peer node. These two features help to support similarity search in two aspects: peer nodes with similar indexing keys would be distributed neighborly, peer nodes with similar indices will connect to each other through routing table. Our evaluation showed that the pService performs considerable Web services discovery efficiency.