AAU Student Projects - visit Aalborg University's student projects portal
A master's thesis from Aalborg University
Book cover


Thor: A tool for detecting energy hotspots in software

Term

4. term

Education

Publication year

2024

Submitted on

Pages

76

Abstract

Previous work have created tools which were able to detect hotspots in software. These tools uses estimation to gather the energy consumption of the different components of a software system. We design and develop a tool, that we call Thor, which does not make use of estimation to gather the energy consumption, but instead uses Intel's Running Average Power Limit to gather the energy consumption. Thor uses static instrumentation and has support for C# and JavaScript with the possibility to extend support to more programming languages. Thor's ability to detect hotspots is tested using two tests. The system used to test Thor, is running the Ubuntu Server operating system. For one of the tests we use an existing method for detecting hotspots, the results from which we use to compare with the results generated by Thor. From the results, we found that, for the first test, Thor was able to detect the hotspots which the existing method found and the reported energy usage was similar. The second test showed Thor in a more complicated scenario where it also detected the hotspots.

Previous work have created tools which were able to detect hotspots in software. These tools uses estimation to gather the energy consumption of the different components of a software system. We design and develop a tool, that we call Thor, which does not make use of estimation to gather the energy consumption, but instead uses Intel's Running Average Power Limit to gather the energy consumption. Thor uses static instrumentation and has support for C# and JavaScript with the possibility to extend support to more programming languages. Thor's ability to detect hotspots is tested using two tests. The system used to test Thor, is running the Ubuntu Server operating system. For one of the tests we use an existing method for detecting hotspots, the results from which we use to compare with the results generated by Thor. From the results, we found that, for the first test, Thor was able to detect the hotspots which the existing method found and the reported energy usage was similar. The second test showed Thor in a more complicated scenario where it also detected the hotspots.