Efficient and Transparent Web-Services Selection

Web services technology standards enable description, publication, discovery of and binding to services distributed towards the Internet. However, current standards do not address the service selection issue : how did a consumer select the service that matches its functional (e.g. operations’ semantics) and non-functional (e.g. price, reputation, response time) properties ? Most projects advocate automatic selection mechanism, advising adaptation or modification of the web-services model and its entities (UDDI, WSDL, Client, Provider). These proposals also do not take advantage of distributed-systems’ state of the art, mainly with respect to the collection and the dissemination of services’ QoS. This paper presents an extension of the initial model that permits automatic service selection, late binding and collection of metrics that characterize the quality of service. The extension consists on a web-service access infrastructure, made of web service proxies and a peer to peer network of QoS metrics repository (the proposal does not impose modification on UDDI registries or services). The proxies interact with common UDDI registries to find suitable services for selection and to publish descriptions. They collect QoS metrics and store them on a p2p network.