Feature interactions induced by data dependencies among entity components

Many large-scale enterprise information systems are data centric, that is, a unique core database is shared among almost all subsystems in an enterprise. When 'developing a new subsystem or upgrading an existing one, the developers have to investigate data dependencies among the subsystems carefully and thoroughly. However, since the core database is very large and the subsystems are usually developed in a parallel, independent and incremental manner, it is very complex, error-prone and time-consuming to detect and analyze data dependencies. As a result, the subsystems may not work well due to some undesired interactions induced by data dependencies, which is similar to feature interactions resulting from resource competition in telecommunications. This paper proposes a middleware-based approach to the automatical detection of data dependencies and potential conflicts in a large-scale enterprise information system, which is demonstrated on J2EE (Java 2 Platform Enterprise Edition). Firstly, we present a case about data dependencies and conflicts in two blueprint applications. Secondly, we extend PKUAS, a J2EE application server, to collect all information about the relationships between the deployed applications and databases and then identify all data dependencies. Thirdly, we define four criteria to detect potential conflicts induced by data dependencies. Finally, we apply the approach to the case of two blueprint applications for automatically detecting data dependencies and conflicts. ? 2005 The authors. All rights reserved.EI