MT-EA4Cloud: A Methodology For testing and optimising energy-aware cloud systems

Abstract

Currently, using conventional techniques for checking and optimising the energy consumption in cloud systems is unpractical, due to the massive computational resources required. An appropriate test suite focusing on the parts of the cloud to be tested must be efficiently synthesised and executed, while the correctness of the test results must be checked. Additionally, alternative cloud configurations that optimise the energetic consumption of the cloud must be generated and analysed accordingly, which is challenging.To solve these issues we present MT-EA4Cloud, a formal approach to check the correctnessfrom an energy-aware point of view -of cloud systems and optimise their energy consumption. To make the checking of energy consumption practical, MT-EA4Cloud combines metamorphic testing, evolutionary algorithms and simulation. Metamorphic testing allows to formally model the underlying cloud infrastructure in the form of metamorphic relations. We use metamorphic testing to alleviate both the reliable test set problem, generating appropriate test suites focused on the features reflected in the metamorphic relations, and the oracle problem, using the metamorphic relations to check the generated results automatically. MT-EA4Cloud uses evolutionary algorithms to efficiently guide the search for optimising the energetic consumption of cloud systems, which can be calculated using different cloud simulators.

Tools, techniques, and trends in sustainable software engineering: A critical review of current practices and future directions

“…Energy-aware testing involves evaluating software systems for their energy consumption under various usage scenarios (Cañizares et al, 2020). Strategies include workload characterization, which involves identifying typical usage patterns and user behaviors, and energy profiling, which measures energy consumption during software execution.…”

A Replication Study on Predicting Metamorphic Relations at Unit Testing Level

“…MT has been demonstrated to be an effective technique for testing in a variety of application domains, e.g., autonomous driving [12], [13], cloud and networking systems [14], [15], bioinformatic software [16], [17], scientific software [18], [19]. However, the efficacy of MT heavily relies on the specific MRs employed.…”

Using Source Code Metrics for Predicting Metamorphic Relations at Method Level

“…Several studies have shown MT as a strong technique for testing the "non-testable programs" where an oracle is unavailable or too difficult to implement [12]- [15]. Also, MT has been demonstrated to be an effective technique for testing in a variety of application domains, e.g., autonomous driving [16], [17], cloud and networking systems [18], [19], bioinformatic software [20], [21], scientific software [22], [23]. However, the efficacy of MT heavily relies on the specific MRs employed.…”