Seamless service composition in pervasive computing environments

In pervasive computing environments, the vision of disappearing technology will become a reality when features available on a number of devices can be utilized in a cooperative manner to accomplish user tasks. Features available from various devices can be exported as services and through service composition they can be combined to work cooperatively toward achieving user tasks. In the past, service composition has been treated as an extended service discovery mechanism where multiple services are discovered with an emphasis on achieving coordination among discovered services. Due to the dynamisms involved in pervasive computing, it is often required to dynamically construct complex services using the available basic services. Traditional service composition mechanisms fall short of constructing such compositions to create required services. In this dissertation, a novel scheme for service description, aggregation and composition called Seamless Service Composition (SeSCo) has been proposed. SeSCo is based on a graph model and is capable of dynamically weaving complex services. With the help of a novel technique called latching, SeSCo is supported by a hierarchical service overlay designed with the intention of exploiting the heterogeneity in pervasive computing environments. The hierarchical overlay is extensively utilized in supporting service related operations such as discovery, composition and execution of services. The hierarchical overlay is used to (i) maintain service information, (ii) resolve service requests, (iii) support resource poor devices in their operation, and (iv) provide seamless support to composite service sessions. While there exist a number of schemes for building and operating service environments, there are no metrics to compare and evaluate service composition schemes. To address this issue, we define various metrics for characterizing and assessing service oriented schemes in pervasive computing environments. SeSCo has been integrated with our pervasive computing middeware called PICO (Pervasive Information Community Organization). Prototypes have been built to validate SeSCo and the results have also been substantiated through experimental evaluation.