An Intelligent Procurement Marketplace for Web Services Composition Megha Mohabey, Y. Narahari Indian Institute of Science, Bangalore, India megha,hari@csa.iisc.ernet.in Sudeep Mallick, P. Suresh, S.V. Subrahmanya Infosys Technologies, Bangalore, India, sudeepm,suresh p01,subrahmanyasv@infosys.com Abstract This paper presents an intelligent procurement market- place for finding the best mix of web services to dynami- cally compose the business process desired by a web ser- vice requester. We develop a combinatorial auction ap- proach that leads to an integer programming formulation for the web services composition problem. The model takes into account the Quality of Service (QoS) and Service Level Agreements (SLA) for differentiating among multiple ser- vice providers who are capable of fulfilling a functionality. An important feature of the model is interface aware com- position. 1 Introduction Mohabey and co-authors have recently presented a com- binatorial auction approach to QoS-aware dynamic web ser- vices composition [4]. Their approach considers not only stand-alone web services but also composite web services to be a part of a business process, as compared to other ap- proaches that consider only stand-alone services for com- position of business process. Such an approach allows the WSPs to provide composite web services at a price and quality that can be different from that of constituent stand- alone services. Moreover, their approach allows representa- tion of scenarios where different WSPs have different num- ber of stages/steps for achieving the same task and they can expose the services at different points. The combinatorial auction leads to an IP formulation for the web services com- position problem. The current work builds upon and extends the above work in the following ways: • The current approaches do not consider the problem of interface matching. Gao et al [3] present an ap- proach for interface aware composition, however, they represent the problem as a weighted multistage graph, which does not take into account global constraints and preferences set by the user (eg. budget constraints). We present a novel way of interface aware composi- tion by means of adding simple constraints to the IP formulation of web service composition problem. • The current approaches [6, 1, 3] consider only QoS and cost for differentiating among various WSPs, however, the SLA guarantees given by various WSPs are also a crucial factor in selecting the web services. When multiple service providers are capable of fulfilling a functionality, our model differentiates on the basis of not only QoS but also the SLA guarantees provided by various WSPs. The rest of the paper is organized as follows. Secion 2 describes the modeling of three important features, namely interface matching, service level agreements, and presence of exogenous service providers. We provide the mathe- matical formulation of the constraints arising due to these features. In Section 3, we briefly describe some numerical experiments to show the effect of modeling these features. The notation for this paper is described in Table 1. 2 Features of the Intelligent Procurement Marketplace 2.1 Interface Matching The services offered by different WSPs may not be inter- operable, if they have incompatible input output interfaces. Service interface is used to transfer control and data mes- sages between web services. When performing the same task, different service implementations may make different data structures and classes, which result in heterogeneities among different interfaces [3]. For example in Figure 1, there are three WSPs capable of providing functionality for the tasks A, B, C, however the interfaces are different. The edges represent compatible interfaces. Thus B 2 can process output of A 1 but not A 3 . We represent different interfaces by l i s. Thus a typical bid structure would be as follows: < index, (S 1 ={t 1 },q 1 ,q 2 ,...,q n ,c(S 1 ),P (S 1 ), L i (S 1 ), L o (S 1 )), (S 2 ={t 2 },q 1 ,q 2 ,...,q n ,c(S 2 ),P (S 2 ), L i (S 2 ), 2007 IEEE/WIC/ACM International Conference on Web Intelligence 0-7695-3026-5/07 $25.00 © 2007 IEEE DOI 10.1109/WI.2007.145 551 2007 IEEE/WIC/ACM International Conference on Web Intelligence 0-7695-3026-5/07 $25.00 © 2007 IEEE DOI 10.1109/WI.2007.145 551 2007 IEEE/WIC/ACM International Conference on Web Intelligence 0-7695-3026-5/07 $25.00 © 2007 IEEE DOI 10.1109/WI.2007.145 551 2007 IEEE/WIC/ACM International Conference on Web Intelligence 0-7695-3026-5/07 $25.00 © 2007 IEEE DOI 10.1109/WI.2007.145 551 Authorized licensed use limited to: INDIAN INSTITUTE OF SCIENCE. Downloaded on November 6, 2008 at 02:23 from IEEE Xplore. Restrictions apply.