One of the programming models that has been developing the most in recent years is Function as a Service (FaaS). The growing concern over data centre energy footprints has driven sustainable software development. In serverless applications, energy consumption depends on the energy consumption of the application’s functions. However, measuring energy proves challenging, and the results’ variability complicates optimisation efforts at runtime. This article addresses this issue by measuring serverless function energy consumption and exploring integration into an optimisation system that selects implementations based on their current energy footprint. For this, we have integrated an energy measurement software into a FaaS system. We have analysed how to properly process the data and how to use them to perform self-adaptation. We present a series of methods and policies that make our system not only capable of detecting variations in the energy consumption of the functions, but it does so taking into account the variability in the measurements that each function may present. Our experiments showcase proper integration in a self-adaptive system, showing a reduction up to 5% in energy consumption due to functions in a test application.

Política de acceso abierto tomada de: https://www.springernature.com/gp/open-research/policies/book-policies

Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech.