Context-aware energy-efficient applications for cyber-physical systems

Abstract

Software systems have a strong impact on the energy consumption of the hardware they use. This is especially important for cyber-physical systems where power consumption strongly influences the battery life. For this reason, software developers should be more aware of the energy consumed by their systems. Moreover, software systems should be developed to adapt their behavior to minimize the energy consumed during their execution. This can be done by monitoring the usage context of the system and having runtime support to react to those changes that impact the energy footprint negatively. Although both the hardware and the software parts of cyber-physical systems can be adapted to reduce its energy consumption, this paper focuses on software adaptation. Concretely, the paper illustrates how to address the problem of developing context-aware energy-efficient applications using a Green Eco-Assistant that makes use of advanced software engineering methods, such as Dynamic Software Product Lines and Separation of Concerns. The main steps of our approach are illustrated by applying them to a cyber-physical system case study.

Statically Analyzing the Energy Efficiency of Software Product Lines

“…Regarding software product lines, several works have been developed with the goal of accurately and, in a lightweight matter, analyzing products independently, and calculate ,for each one, properties such as correctness and performance. Nevertheless, the energy consumption of software products is yet a narrowly explored concept [14,15]. In [14], an initial approach for statically inferring energy estimations was presented.…”

“…In [14], an initial approach for statically inferring energy estimations was presented. The strategy that was followed in [15] was to create an approach to build Dynamic SPLs and, through a dynamic analysis approach (involving continuous usage monitoring), the SPL could adapt itself to avoid using features potentially leading to excessive consumptions. To the best of our knowledge, these are the only research works performed for SPLs that target energy analysis.…”

“…Although several techniques have been proposed to measure, reason about, and improve the energy efficiency within software and language engineering [9][10][11][12][13], only recently has there been an effort towards addressing this issue in SPLs [14,15]. In a nutshell, a SPL development approach was presented in [15], which was mainly focused on building a DSPL that can adapt itself with energy-saving capacities, through establishing a relation between variability features and their energy impact and continuous monitoring the context on which the SPL is used. Regarding energy analysis of all products within a SPL, we have proposed in [14] the first approach to explore static analysis to reason about such products, in order to obtain for each one accurate energy consumption estimations.…”

Energy efficient adaptation engines for android applications

“…However, Hasan et al [52] measure the energy consumption of the Java collections classes by creating energy profiles, and demonstrate that choosing an energy efficient collection can improve energy consumption of Android application by as much as 38%. Similar analysis of energy impact have been performed over other specific functionalities such as logging [39] , compression [56] , the HTTP protocol [57] , testing frameworks [58] , or the operationalization of quality attributes [7] . Raising the abstraction level of the refactoring, Morales et al [54] analyze the impact of anti-patterns on the design of Android applications, and propose EARMO, an anti-pattern correction approach based on evolutionary multi-objective techniques that accounts for energy consumption when refactoring mobile anti-patterns.…”

Supporting the evolution of applications deployed on edge-based infrastructures using multi-layer feature models

“…Apart from evaluating the time required by our modules to return a solution, we also planned to measure their energy consumption associated using hardware energy meters [18,20]. Nonetheless, it has not been possible as evaluation was done during the teleworking restrictions imposed by the COVID-19 global pandemic.…”