Performance analysis framework for large software-intensive systems with a message passing paradigm

• Authors: Christian Del Rosso
• Venue: Proceedings of the 2005 ACM symposium on Applied computing
• Year: 2005

The launch of new features for mobile phones is increasing and the product life cycle symmetrically decreasing in duration as higher levels of sophistication are reached. Therefore, the optimization of resources is particularly important in embedded systems where CPU power and memory space are limited. In this context, performance engineers must be able to predict and analyze the performance of the software architecture in order to support its evolution and its new requirements. In this paper I describe a framework for the analysis of the performance of a software architecture where the architectural elements communicate using message based communication services. Using instrumentation traces I have extracted the run-time events I considered significant for the study. I have created a set of architectural views to reconstruct the dynamic and the static views of the architecture. Understanding the connections and the relationships between them guide the performance analyst to a clear comprehension of how the architecture works and subsequently how it can be optimized. The performance analysis framework described constitute an essential set. The next challenge is to enhance the integration of the tools and the synchronization of the views and to facilitate the entry barrier to novice performance engineers.