Efficient Web Services Discovery and Composition

As an emerging cross discipline area for distributed computing, Service-Oriented Computing (SOC) paradigm promises to bridge the gap between Business Services and IT Services, enable technology to help people perform business processes more efficiently and effectively, and enable businesses and organizations to collaborate globally through standard services. With the rapid development and popularity of Internet and e-commerce, business organizations are investigating ways to expose their current software components into web services so as to make use of distributed service computing resources. Business organizations are also investigating ways on how to incorporate services running on different platforms and hosted by service providers outside of their boundaries into more complex, orchestrated services. This conceptually new approach brings multiple issues to researchers. As the number of web services and SOC applications increases, there is a growing need for mechanisms for discovering services efficiently. Effective mechanisms for web services discovery and ranking are critical for organizations to take advantage of the tremendous opportunities offered by web services, to engage in business collaborations and service compositions, to identify potential service partners, and to understand service competitors and increase the competitive edge of their service offerings. Apart from functional specification required by users during service discovery, some non-functional requirements such as Quality of Service (QoS) and transactional properties are also major concerns, because users need to select and compose web services not only according to their functional requirements but also to their transactional properties and QoS characteristics defined using a quality model. On one hand, in service-oriented environments multiple web services may provide similar functionality, but may offer different non-functional properties (e.g., price). When selecting existing web services from web service communities to generate composite services, the number of composed services may be very large. Therefore, it is a challenge on how to select appropriate web services to satisfy users’ global QoS requirements based on a set of given QoS preferences. On the other hand, as web services operate in a highly dynamic distributed environment and interact with each other, the possibility of unexpected behavior is high. The unexpected behavior from an individual service may bring negative impact on all the component services in the composition, even lead to failure of the running of the composite service. Thus we need a mechanism that provides transactional support to service composition to ensure the overall consistent and reliable execution of business processes. In this thesis, we investigate the problem of efficient web services discovery and composition in service oriented environments. Firstly, we present an efficient IR-Style mechanism for discovering and ranking web services automatically, given a textual description of desired services. We introduce the notion of preference degree for a web service, and suggest relevance and importance as two desired properties for measuring its preference degree. Also, various algorithms are given to obtain service relevance and importance. The key part for computing service importance is a new schema tree matching algorithm, which catches not only structures, but even better semantic information of schemas defined in web services. Moreover, we develop an approach to identify associations between web-service operations based on service operations matching. This approach uses the concept of attribute closure to obtain sets of operations. Each set is composed of associated web-service operations. Experimental results show the proposed IR-style search strategy is efficient and practical. Secondly, we propose a novel QoS model for performing flexible service selection in web service composition with QoS constraints. The key idea of the model is to relax users’ QoS constraints and try to find the most possible services satisfying users’ QoS requirements. Based on the proposed QoS framework, we develop various algorithms for making service selection on individual and composite services. We also introduce a top-k ranking strategy to reflect users’ personalized requirements. We evaluate the performance of the model in terms of time and space cost. Finally, we propose a strategy to verify at design time whether a service composition can be implemented by a set of selected web services according to the transactional requirements specified by a user. We define transactional properties for component web services and analyze the dependencies among them. Next, based on the Automata theory, we model the transactional behavior of both component web services and composite services as transition systems, and use Accepted Termination States (ATS) to describe the correct execution of composite services. Then, by analyzing the message exchanges between component services of a composite service we present a novel algorithm to verify all possible state transitions of the composite service to ensure consistent termination states according to the ATS specified by the user. In addition, for the cause of efficiency we use Temporal Logic to describe the transactional properties of a composite service, and then we employ SPIN, an automata-based model checker to carry out automated verification of temporal logic properties on the composite service.