Detection of SOA Patterns

The rapid increase of communications combined with the deployment of large scale information systems lead to the democratization of Service Oriented Architectures (SOA). However, systems based on these architectures (called SOA systems) evolve rapidly due to the addition of new functionalities, the modification of execution contexts and the integration of legacy systems. This evolution may hinder the maintenance of these systems, and thus increase the cost of their development. To ease the evolution and maintenance of SOA systems, they should satisfy good design quality criteria, possibly expressed using patterns. By patterns, we mean good practices to solve known and common problems when designing software systems. The goal of this study is to detect patterns in SOA systems to assess their design and their Quality of Service (QoS). We propose a three steps approach called SODOP (Service Oriented Detection Of Patterns), which is based on our previous work for the detection of antipatterns. As a first step, we define five SOA patterns extracted from the literature. We specify these patterns using "rule cards", which are sets of rules that combine various metrics, static or dynamic, using a formal grammar. The second step consists in generating automatically detection algorithms from rule cards. The last step consists in applying concretely these algorithms to detect patterns on SOA systems at runtime. We validate SODOP on two SOA systems: Home-Automation and FraSCAti that contain respectively 13 and 91 services. This validation demonstrates that our proposed approach is precise and efficient.