In order for Web services to be able to fully exploit the advantages of dynamic binding, requesting agents must be given a way to select the most ”convenient” service at invoke time. Current techniques do not address the problem of selecting services that best fit a requester’ needs. In this paper, we propose an extension of the UDDI specification in order to provide and exploit predictions about the behavior of Web services. Such values/parameters are expressed in terms of availability, reliability and completion time. The computed predictions are used by requesting agents to implement an effective dynamic service selection. A prototype, named eUDDIr, implementing such extension has been developed. Our proposal is particularly suited for scenarios where requestors do not wish to explicitly deal with QoS aspects, or in case provider agents have no convenience in building up the infrastructure for guaranteed QoS, but aim to provide services of good quality to their customers. The adoption of eUDDIr can effectively improve the service requestors ”satisfaction”, in particular when they are involved in a Web services composition process.
Ambrosi, E., Bianchi, M., Gaibisso, C., Gambosi, G., Lombardi, F. (2005). Extending the UDDI API for service instance ranking. ??????? it.cilea.surplus.oa.citation.tipologie.CitationProceedings.prensentedAt ??????? International Symposium on Web Services and Applications, ISWS 2005, Las Vegas, Nevada, USA.
Extending the UDDI API for service instance ranking
GAMBOSI, GIORGIO;
2005-01-01
Abstract
In order for Web services to be able to fully exploit the advantages of dynamic binding, requesting agents must be given a way to select the most ”convenient” service at invoke time. Current techniques do not address the problem of selecting services that best fit a requester’ needs. In this paper, we propose an extension of the UDDI specification in order to provide and exploit predictions about the behavior of Web services. Such values/parameters are expressed in terms of availability, reliability and completion time. The computed predictions are used by requesting agents to implement an effective dynamic service selection. A prototype, named eUDDIr, implementing such extension has been developed. Our proposal is particularly suited for scenarios where requestors do not wish to explicitly deal with QoS aspects, or in case provider agents have no convenience in building up the infrastructure for guaranteed QoS, but aim to provide services of good quality to their customers. The adoption of eUDDIr can effectively improve the service requestors ”satisfaction”, in particular when they are involved in a Web services composition process.Questo articolo è pubblicato sotto una Licenza Licenza Creative Commons
