SIM: An Automated Approach to Improve Web Service Interface Modularization

Service interface structure is of primary importance in SOA to ensure best practice of third-party reuse. One of the key factors for deploying successful services is assuring an adequate interface structure. However, a common bad service design practice is to place semantically unrelated operations in a single interface. This poor design practice typically result in a system which is difficult to comprehend, maintain and evolve providing low performance and reusability. To address this problem, we present an automated approach, SIM, to support service developers improve the quality of their interface modularization. Our approach analyzes structural and semantic relationships among the operations exposed in a service interface to identify chains of strongly related operations. The identified operation chains are used to define new interfaces with higher cohesion and better usability. We empirically evaluate our approach on a benchmark of 22 realworld Web services, provided by Amazon and Yahoo. The obtained results show that the produced interfaces are (i) able to improve the service design quality, and (ii) recognized as 'useful' from developers point of view in improving their service design. Additionally, we found that SIM significantly outperforms a recent state-of-the-art approach.