State Specialization in a Service Discovery Ontology: A Financial Services Business Grid David Bell Department of Information Systems and Computing Brunel University UK David.Bell@brunel.ac.uk Simone A. Ludwig School of Computer Science Cardiff University UK Simone.Ludwig@cs.cardiff.ac.uk Mark Lycett Department of Information Systems and Computing Brunel University UK Mark.Lycett@brunel.ac.uk Abstract Investment Banking requires a diverse set of supporting systems in order to operate in a range of markets from corporate mergers to trading options on weather. The challenge to this community is the ability to adapt to new business requirements in an effective manner, utilizing their network of capabilities in a flexible and dynamic way. A semantic approach to capability discovery can combine many strategic perspectives in a pragmatic yet easily usable form. The use of richer explicit knowledge, that is system readable, provides the basis for discovering capabilities on this exemplar Business Grid - "the grid of services". Current research of semantic capability description in the Grid community has tended to focus on resource selection rather than service discovery. This research explores the practical usage of Grid services in the financial market sector. The approach demonstrates the need for distributed and phased semantic service discovery (with capabilities described and stored in a dynamic ontology). 1 INTRODUCTION Investment banks house a diverse set of systems segregated by product, process or geographical focus. The product range include Foreign Exchange, Interest Rates, Fixed Income (Bonds), Commodities (from Oil to Weather) and Derivatives. Action is then required to bring together capabilities resident in several products, processes or geographical focused systems. The challenge to these organizations is to be able to adapt to new business requirements in an effective way, utilizing their capabilities in a flexible manner. The size of the inventory directs this research away from centralized knowledge engineering and use; and toward strategies for service knowledge segmentation. Focusing ones thinking on process and performance directs this analysis. Grid infrastructure, with its Web Services layer, provides an infrastructure for exposing stateless and stateful organizational capabilities for re-use and novel re-configuration. The logical next step, with a service oriented inventory, is to identify practical methods for the discovery of these capabilities in the support of sporadic business need. Existing methods such as UDDI (Universal Description Discovery and Integration) [1] and MDS (Monitoring and Discovery Service) [2] were discounted due to the limitation of a string matching syntactic approach [3]. We have chosen a semantic approach to service discovery; investigating the practicalities of semantic technology in Financial Services (FS). The experiments explore the performance impact of top down conceptual models and bottom up state models; with results directing a novel phased combination of the two ontology formation mechanisms. The remainder of this paper is organized as follows. In section 2 related work which motivates the research executed is presented. Section 3 shows the financial service architecture and describes the components used. In section 4 the implementation of the prototype is described with the tools used. Section 5 shows the performance graphs for two phases of experimentation. A conclusion is given in section 6 which summarizes the findings and suggests some further improvements. 2 RELATED WORK Many general strategies have been proposed to address the issues of adapting to new business requirements in an effective way, utilizing these capabilities in a flexible manner. Social sciences have proposed organizational learning (OL), with tacit- explicit conversion of knowledge. This focus on knowledge flow, whilst often not computer readable, provides a basis for supporting improved inter silo decision making, clearly highlighted in Nonaka’s knowledge spiral [4]. Alternatively, technology organizations have recently refocused on the use of workflow to orchestrate the execution of capabilities within the system inventory (an example being BPEL4WS [5]). One problem here is that no single