Interacting with Service Compositions

Interaction with web service enabled marketplaces would be greatly facilitated if users were given a high level service request language to express their goals in complex business domains. This could be achieved by using a planning framework which monitors the execution of planned goals against predefined standard business processes and interacts with the user to achieve goal satisfaction. The thesis addresses the problem of service composition executions managed by interacting with the client. The planning architecture accepts high level requests, expressed in XSRL (XML Service Request Language). The planning framework is based on the principle of interleaving planning and execution. This is accomplished on the basis of refinement and revision as new service-related information is gathered from service registries and web services instances, and as execution circumstances necessitate change. The system interacts with the user whenever confirmation or verification is needed. The work is primarily concerned with the problems of composition and monitoring of business processes based on process reference models. XML Service Request Language (XSRL) was proposed to address the issue of service composition by giving the users an explicit control over process executions by describing desired service attributes and functionalities, including temporal and non-temporal constraints between services. Reference model instantiation is planned according to the goals and preferences specified by the user, and an appropriate plan is executed. The algorithms behind are based on the idea of interleaving planning and execution. These algorithms are based on model checking and constraint programming. Research in process monitoring was brought to the definition of a framework where business rules are defined by assertion statements. Assertions are published by the partners involved in the business process. The business process is executed with respect to the specified assertions and, by that, up-to-date business objectives, constraints and new market situations. Algorithms for interleaving planning, monitoring and execution have been implemented by using Java programming language. The implementation uses constraint programming system to satisfy user goals and preferences against reference business processes. The choice for constraint solvers was motivated by the fact that in a web service scenario, users may wish to know why certain solutions are preferred to others. Explanation-based constraint programming is a viable approach to tackling such issues. As a constraint solver external system (Choco) is used. Choco is a Java library for constraint satisfaction problems, constraint programming and explanation-based constraint solving. The evaluation of the language and service composition framework focused on showing the feasibility of the approach. There was implemented a domain generator allowing the tuning of the following parameters: number of states, branching factor, non-determinism rate, maximum directed cycle length, and if the service domain corresponds to a dag or a not. Then there were made several experimentations with different XSRL requests with increasing complexity on different domains. The evaluation showed that the system is able to deal with domains which are considerably larger than the biggest ones seen in practice.