SEEDS: a software engineer's energy-optimization decision support framework

Abstract

Reducing the energy usage of software is becoming more important in many environments, in particular, batterypowered mobile devices, embedded systems and data centers. Recent empirical studies indicate that software engineers can support the goal of reducing energy usage by making design and implementation decisions in ways that take into consideration how such decisions impact the energy usage of an application. However, the large number of possible choices and the lack of feedback and information available to software engineers necessitates some form of automated decision-making support. This paper describes the first known automated support for systematically optimizing the energy usage of applications by making code-level changes. It is e↵ective at reducing energy usage while freeing developers from needing to deal with the low-level, tedious tasks of applying changes and monitoring the resulting impacts to the energy usage of their application. We present a general framework, SEEDS, as well as an instantiation of the framework that automatically optimizes Java applications by selecting the most energye cient library implementations for Java's Collections API. Our empirical evaluation of the framework and instantiation show that it is possible to improve the energy usage of an application in a fully automated manner for a reasonable cost.

Object-Oriented Genetic Improvement for Improved Energy Consumption in Google Guava

“…The GA required approximately 3,500 fitness evaluations to find its best solution (the number of evaluations varied slightly between different evolutionary runs), while the exhaustive search at each variation point required only 105 evaluations. Although using the GA was therefore significantly slower than the approach used in similar work [9], the final result is also significantly better.…”

“…Thus, the results show that the substitutions improve both the energy consumption and the execution time of the class. We further compared the results of our technique to those obtained using an approach used in related work [9]-a separate exhaustive search at each variation point-and found that although the number of fitness evaluations increased using a GA, the performance of the final result was significantly improved. This shows that the variation points within code are not always independent.…”

“…The final experiment ran an exhaustive search independently on each variation point, following the example of Manotas et al [9] but on source code instead of bytecode, before combining each of the substitutions at the end.…”

“…the Liskov Substitution Principle (LSP) [7] as exemplified by subtype polymorphism. In this respect, the closest work we are aware of is that of the SEEDS framework of Manotas et al [9], in which alternative subtypes of container classes are substituted into bytecode in order to minimize power consumption. The search mechanism that is employed in SEEDS is that of a separate exhaustive search at each object allocation location: our approach differs by using a genetic approach to assign subclasses to constructor invocations.…”

Achievements, Open Problems and Challenges for Search Based Software Testing

“…For example, Li et al [80] formulate energy optimisation as the problem of finding the mobile device screen colour choices that minimise energy consumption, while maintaining colour contrast. Monotas et al [83] also define a search space for energy optimisation choices. They currently use an exhaustive search, but plan to extend to full SBSE for scalability to larger search spaces.…”