Performance analysis of UML models using aspect-oriented modeling techniques

Aspect-Oriented Modeling (AOM) techniques allow software designers to isolate and address separately solutions for crosscutting concerns (such as security, reliability, new functional features, etc.) This paper proposes an approach for analyzing the performance effects of a given aspect on the overall system performance, after the composition of the aspect model with the primary model of a system. Performance analysis of UML models is enabled by the "UML Performance Profile for Schedulability, Performance and Time" (SPT) standardized by OMG, which defines a set of quantitative performance annotations to be added to a UML model. The first step of the proposed approach is to add performance annotations to both the primary model and to the aspect model(s). An aspect model is generic at first, and therefore its performance annotations must be parameterized. A generic model will be converted into a context-specific aspect model with concrete values assigned to its performance annotations. The latter is composed with the primary model, generating a complete annotated UML model. By using existing techniques, the complete model is transformed automatically into a Layered Queueing Network (LQN) performance model, which can be analyzed with existing solvers. The proposed approach is illustrated with a case study system, whose primary model is enhanced with some security features by using AOM. The LQN model of the primary system was validated against measurements in previous work. The performance effects of the security aspect under consideration are analyzed in two design alternatives by using the LQN model of the composed system.