Scalable composition of Web services under uncertainty
MetadataShow full item record
Facilitating the assembly of services to form composite services is an important functionality in Service-oriented architecture (SOA). In this dissertation, we focus on the problem of automatically assembling WSs to form compositions that optimize given user preferences. This problem is often referred to as the automated Web service composition problem. Prevalent approaches for automatically composing WSs predominantly utilize planning techniques to achieve the composition. However, classical artificial intelligence (AI) planning based approaches are facing the following issues: (1) They are incapable of capturing the uncertainties of Web service behaviors, (2) It is hard for them to provide process optimization during planning, and (3) Many of them are unable to scale efficiently to large processes. To address these issues, we present a hierarchical decision-theoretic planning framework for composing Web services, called Haley. Compared to classical AI planners, the decision-theoretic planning has the ability to capture the uncertainty inherent in WSs and provide a cost based process optimization. Haley uses symbolic planning techniques that operate directly on first order logic based representations of the state space to obtain the compositions. As a result, it supports an automated elicitation of the corresponding planning domain from WS descriptions and produces a compact domain representation in comparison to classical AI planners. Additionally, it tackles the scalability issue by exploiting the hierarchy found in processes. Our experiments demonstrate that Haley evaluates favorably in comparison to other WS composition approaches. We implement Haley and provide a comprehensive tool suite. The suite accepts WSs described using standard languages such as SAWSDL. It provides process designers with an intuitive interface to specify process requirements, goals and a hierarchical decomposition, and automatically generates BPEL processes, while hiding the complexity of the planning and BPEL from users. Another emerging research topic is automated REpresentational State Transfer (REST)ful WS composition. While automating WSDL/SOAP WS composition has been extensively studied, automated RESTful WS composition is less explored in the research community. As an early effort addressing this problem, this dissertation discusses the challenges of composing RESTful WSs and proposes a formal model for describing and automatically composing RESTful WSs.