Continuous Integration of Performance Model

Abstract

Applying model-based performance prediction requires that an up-to-date Performance Model (PM) is available throughout the development process. Creating such a model manually is an expensive process that is unsuitable for agile software development aiming to produce rapid releases in short cycles. Existing approaches automate the extraction of a PM based on reverse engineering and/or measurements techniques. However, these approaches require to monitor and analyse the whole application. Thus, they are too costly to be applied frequently, up to after each code change. Moreover, keeping potential manual changes of the PM is another challenge as long the PM is regenerated from scratch every time. To address these problems, this paper envisions an approach for efficient continuous integration of a parametrised performance model in an agile development process. Our work will combine static code analysis with adaptive, automatic, dynamic analysis covering updated parts of code to update the PM with parameters, like resource demands and branching probabilities. The benefit of our approach will be to automatically keep the PM up-to-date throughout the development process which enables the proactive identification of upcoming performance problems and provides a foundation for evaluating design alternatives at low costs.