Investigating the Use of Model-Driven Generative Techniques for Middleware Performance Analysis
Source: University of Alabama
The increasing demand for faster availability of new, feature-rich and high performance network-centric services imposes the need for these services to be rapidly developed via composition, configuration and assembly of standards middleware building blocks. However, the decisions on the systems compositions must be validated for performance guarantees at design time, which is addressed in this paper. First, the authors describe the manual process in developing a performance model for a middleware building block and illustrate the challenges in scaling such models for large-scale compositions. Second, they show how Model-Driven Development (MDD) can play a key role in addressing these scalability challenges. They describe a MDD performance evaluation framework they are developing. Third, they illustrate the framework's capabilities on a simple case study demonstrating the viability of this approach.