The hunt for the guzzler: Architecture-based energy profiling using stubs

Abstract

Context: Software producing organizations have the ability to address the energy impact of their software products through their source code and software architecture. In spite of that, the focus often remains on hardware aspects, which limits the contribution of software towards energy efficient ICT solutions. Objective: No methods exist to provide software architects information about the energy consumption of the different components in their software product. The objective of this paper is to bring software producing organizations in control of this qualitative aspect of their software. Method: To achieve the objective, we developed the StEP Method to systematically investigate the effects of software units through the use of software stubs in relation to energy concerns. To evaluate the proposed method, an experiment involving three different versions of a commercial software product has been conducted. In the experiment, two versions of a software product were stubbed according to stakeholder concerns and stressed according to a test case, whilst energy consumption measurements were performed. The method provided guidance for the experiment and all activities were documented for future purposes. Results: Comparing energy consumption differences across versions unraveled the energy consumption related to the products’ core functionality. Using the energy profile, stakeholders could identify the major energy consuming elements and prioritize software engineering efforts to maximize impact. Conclusions: We introduce the StEP Method and demonstrate its applicability in an industrial setting. The method identified energy hotspots and thereby improved the control stakeholders have over the sustainability of a software product. Despite promising results, several concerns are identified that require further attention to improve the method. For instance, we recommend the investigation of software operation data to determine, and possibly automatically create, stubs.