Workshop on automatic service composition

Abstract

In Service-Oriented Architecture (SOA), applications can be assembled by a set of existing distributed services spanning across organizations and platforms. SOA increases reuse and cooperation among services. Therefore, applications can be built in a rapid way with relatively low cost. With the increasing adoption of SOA solutions to application integration, the ability to compose services efficiently and effectively becomes more and more important. A lot of efforts have been devoted to service composition. Service composition integrates existing reusable services together as a process to fulfill business requirements, and exposes the process as a service (i.e., composite service). Specifically, a composition process involves the following five steps: 1. Specify business requirements: A requirement (e.g., purchase a book online) for composing services could be a declarative expression using formal or informal languages, an abstract process model (e.g., a business process model), or design models (e.g., UML diagrams). 2. Provide service description: Service description specifies the capability of services, such as inputs/outputs, exceptions, functional and non-functional description. 3. Discover relevant services: Service discovery is the process of locating a service that meets a certain searching criteria. For example, we search for an appropriate service to offer forecasting service. The searching criteria for the service discovery could be specified using keywords, such as weather forecast. 4. Generate composition process model: A composition process model contains a set of selected services, control flows and data flows among those services. Standards, such as BPEL (Business Process Execution Language), are used to formally specify process models. A business requirement usually cannot be fulfilled by a single service. Various services are integrated using composition process models. 5. Evaluate composition results: It is quite common that many returned services have the same or similar functionalities. It leads to more than one composition result to fulfill the business requirements. The different composition results can be evaluated to select the best one that meets the functional and nonfunctional requirements.