Authors: Pablo Rodriguez-Mier, Carlos Pedrinaci, Manuel Lama, and Manuel Mucientes

Abstract—Intensive research efforts have been focused on solving the automatic discovery or the automatic composition of Web services. However, research and development on both fields has for the most part remained disconnected, without considering the existing overlap between them. This has led to work replication and prevented the adequate integration of discovery and composition engines which are key components at the core of most Service Oriented Architectures. In this paper we present a framework based on a theoretical analysis of service composition in terms of its dependency with service discovery, and a reference implementation of this framework on the basis of two pre-existing separate components, namely iServe and ComposIT. This reference implementation has been used to empirically study the impact of discovery and matchmaking on service composition, and we have provided three different configurations with varying performance. The empirical analysis proves the scalability and flexibility of our proposal and provides insights on how integrated composition systems can be designed in order to achieve good performance in real scenarios, where service registries and composition frameworks are likely to be distributed and controlled by diverse organisations.

The purpose of this document is to extend the results obtained using the Web Service Challenge 2008 datasets with the most recent version of this challenge (Web Service Challenge 2009-2010). The main reason behind including only the results with the WSC'08 datasets is that the WSC'09-10 competition is focused on Quality-Of-Service (QoS) optimisation, and therefore better results can be achieved by developing concrete techniques for filtering services by their QoS. Moreover, this difference makes our results not comparable with the other approaches since we are providing semantic compositions just optimising the composition length and the number of services ignoring QoS.

We are currently working towards extending the proposed framework with QoS. Meanwhile, we think these results can be also interesting to compare the scalability of different composition frameworks.

The datasets used in the 2009 and 2010 are the same datasets. The difference between both competitions are the rules to evaluate the results. However, since we are not considering QoS and our results are not comparable we are not affected by this criterion.