Details

Autor(en) / Beteiligte

• Carção, Tiago Alves

Titel

Spectrum-Based Energy Leak Localization

Ort / Verlag

ProQuest Dissertations Publishing

Erscheinungsjahr

2014

Link zum Volltext

• ProQuest

Quelle

ProQuest Dissertations & Theses A&I

Beschreibungen/Notizen

• For the past few years, we have begun to witness an exponential growth in the inform ation and communication technologies (ICT) sector. While undoubtedly a milestone, all of this occurs at the expense of high energy costs needed to supply servers, data centers, and any use of computers. Associated with these high energy costs is the emission of greenhouse gases. These two issues have become major problems in society. The ICT sector contributes to 7% of the overall energy consumption, with 50% of the energy costs of an organization being attributed to the information technology (IT) departments. Most of the measures taken to address the high level of energy consumption have been on the hardware side. Although is the hardware that does consume energy, it is the software that operates that hardware. As a consequence, the software is the main responsible for the energy consumed by the hardware, very much like a driver that drives/operates a car influences drastically the fuel consumed by the car. This dissertation proposes and implements a methodology to analyze the software energy consumption. This methodology relates energy consumption to the source code of a soft ware application, so that software developers are aware of the energy footprint that he/she is creating with his/her application. The proposed technique interprets abnormal energy consumption as software faults, and adapts a well-known technique for locating faults on programs’s source code, to locate “energy faults”, that we name as “energy leaks”. This methodology has been fully implemented in a software framework that monitors the energy consumed by a software program and identifies its energy leaks, given its source code. Moreover, a list of problematic parts of the code is produced, thus, helping software developers identifying energy faults on their source code. We validate our findings by showing that our methodology can automatically find energy leaks in programs for which such leaks are known. With this results, one intends to provide help to the development phase and to gener ate more energy efficient programs that will have less energy costs associated with, while supporting practices that promote and contribute to sustainability.