Towards Automatic Service Composition within ARGUGRID

Workflow systems play an important role in service-oriented computing as they provide an intuitive mechanism for orchestrating the execution of remote services. Constructing new workflows from raw services however, is not always a straight-forward task. It requires resolving many decisions including locating available services, determining which of them match the user requirements and also deciding how to compose them together into valid applications. Workflow construction activities can be simplified, or possibly even automated, by using a combination of semantic annotations and delegation of the decision making tasks to intelligent agents. Two key challenges arise when developing a practical system that attempts to address this vision. The first is elaborating the key properties of service workflows and the decisions that agents are required to assist in. This information is needed for designing and implementing the internals of the agent mind. The second is designing and implementing the run-time interfaces between the workflow system and the agent system to enable the exchange of information between them. This chapter describes our experience in addressing these two challenges and in developing a framework that simplifies the workflow creation process within the ARGUGRID project.