Pattern-Based Modifiability Analysis of EJB Architectures

Abstract

Over the last years, several techniques to evaluatemodifiability of software architectureshave been developed. One of such techniques is change impact analysis (CIA), which aids developers in assessing the effects of change scenarios on architectural modules. However, CIA does not take into account the pattern structures behind those modules. In architectural frameworks, such as the Enterprise Java Beans (EJB) architecture, the use of patterns is a key practice to achieve modifiability goals. Although patterns can be easily understood individually, when an application combines several pattern instances the analysis is not straightforward. In practice, many EJB designsare assessed in an ad-hoc manner and relying on the developers' experience. A way of dealing with this problem is through the integration of modifiabiliy analysis models and patterns. We propose a knowledge-based approach that explicitly links the EJB patterns to a scenario-based analysis for multi-tier architectures. Specifically, we have developed a modifiability reasoning framework that reifies the EJB patterns present in a given design solution and, for a set of predetermined scenarios, the reasoning framework identifies which architectural elements can be affected by the scenarios. The reasoning framework outputs metrics for each of the scenarios regarding specific EJB tiers. The main contribution of this approach is that assists developers to evaluate EJB alternatives, providing quantitativeinformation about the modifiability implications of their decisions. A preliminary evaluation has shown that the the reasoning framework is viable to analyze EJB designs.