Discovery of Web Services in a Federated Registry Environment Kaarthik Sivashanmugam, Kunal Verma, Amit Sheth Large Scale Distributed Information Systems (LSDIS) Lab Department of Computer Science, University of Georgia Athens, GA 30602 {verma, amit}@cs.uga.edu Abstract The potential of a large scale growth of private and semi-private registries is creating the need for an infrastructure which can support discovery and publication over a group of autonomous registries. Recent versions of UDDI have made changes to accommodate interactions between distributed registries. In this paper, we discuss METEOR-S Web Service Discovery Infrastructure, which provides an ontology based infrastructure to provide access to registries divided based on business domains and grouped into federations. We also discuss how Web service discovery is carried out within a federation. Keywords: Web services Discovery, peer to peer, Decentralized UDDI 1. Introduction There has been a significant change in focus of the vision of UDDI [UDDI, 2002] since its inception. This was evident in the release of version 3 [UDDI, 2003], which has several new features to augment the centralized paradigm of UBR to facilitate interaction between the UBR and private and semi-private registries. The current search facilities offered by the latest version of UDDI do not offer any special features for finding Web service registries. As a result, it is assumed that Web service clients have prior knowledge of the location of the registries. In this paper, we present our implementation of a peer to peer network of private semi-private and public UDDI registries, which allows transparent access to other registries based on registry federations or domains. We use an ontology based approach to classify registries and locate them based on the users’ requirements. Let us consider the following scenario which illustrates the benefits of private registries and having the ability to interact with other private registries. We can imagine a manufacturer that maintains a private registry to maintain details about its suppliers and other partners. Now consider a case when its suppliers are unable to meet its demands either due to adverse circumstances or large orders, and the manufacturer has to locate other suppliers. Due to trust issues, the manufacturer may not want to search the UBR and find just any supplier. It may however want to request its partners or competitors for references about the trusted suppliers or it may want to contact a marketplace to find similar services. Assuming that partners maintain similar private registries, this process can be automated by forming registries federations, where the registry owners give only members of the federation access to their registries. Forming a federation of registries will allow businesses to share their data while maintaining their privacy. In this paper, we leverage the METEOR-S Web Service Discovery Infrastructure (MWSDI) [Verma et al., 2004] for providing transparent access to private and public Web service registries. The focus of this paper is the creation of registry federations. We present a discussion of registry federations and characterize them in the dimensions of distribution, autonomy and heterogeneity. In order provide efficient access to the registries we store semantic metadata Web service registry community in the Extended Registries Ontology (XTRO). We also discuss the different kinds of querying possible using our infrastructure. This work was done as part of the METEOR-S project, which aims to create an infrastructure for complete lifecycle management of Semantic Web processes. We briefly describe the need for decentralization in section 2. The technical details of MWSDI are described in section 3. The classes and relationships of the XTRO are shown in section 4. In section 5, we discuss and analyze Registry Federations. In section 6, we have discussed implementation details in distributed registries with the help of tModels. Service publication and discovery have been discussed in sections 7 and 8. Section 9 discusses related work. Conclusions and future work are mentioned in section 10. 2. Need for Decentralization One of the main reasons that the UDDI specifications decided to acknowledge replication model for data in registries instead of distribution is to enable inspection of data in UDDI along multiple perspectives. There can be potentially several facets to distribute data in UDDI, such as those related to geographical location, nature of registered services, business functionality, technical specifications and so on. The data partitioning on the other dimension could be hierarchical or non-hierarchical. With such distribution architecture, it should be possible notonly to locate appropriate registry (by processing all kinds of data distribution facets and data partitioning criteria) but also to aggregate search results from different candidate registries. Replication was chosen in UDDI because creating a scalable model for distribution of data is inherently difficult. However, in version 3 of the UDDI specifications, the need for data partitioning and affiliation among registries have been acknowledged. In MWSDI, we advocate data distribution (as opposed to data replication) and support any kind of data distribution among multiple registries1. The data partitioning criteria is stored in the Extended Registries Ontology (XTRO) in MWSDI. 3. MWSDI In this section, we briefly describe the conceptual foundations and implementation of MWSDI which provides the infrastructure for the work presented in this paper. Detailed descriptions of the architecture, implementation and protocols can be found in [Verma et al., 2004]. The aim of MWSDI is to provide clients with an efficient publication and discovery mechanism in a multi registry environment. We use semantic metadata stored in the XTRO to identify appropriate registries and direct the queries to them. Each time, a new registry is added to MWSDI, the XTRO is updated with the relevant details of the new registry. We show the interaction of MWSDI, registries and clients in Figure 1. Figure 1: Interaction of MWSDI with clients and registries 1 Our prototype implementation is tested with non-hierarchical distribution of data. Data replication is also supported using the “replicateOf” relationship in XTRO. We have implemented MWSDI as a peer to peer network. Based on the different