Architectural Tactics to Improve the Environmental Sustainability of Microservices: A Rapid Review
Xingwen Xiao
TL;DR
This rapid review identifies six architectural tactics to enhance the environmental sustainability of microservices, emphasizing resource efficiency as the most studied dimension. It synthesizes evidence from 22 peer-reviewed studies, categorizing tactics by sustainability aspect (energy, resource, carbon) and deployment context, and detailing measurement approaches and toolchains used to assess effectiveness. The findings show predictive autoscaling and elastic containers as prominent tactics, with common experimentation environments built on Docker/Kubernetes and standard monitoring stacks, while greenability and direct carbon measurements remain underexplored. The study highlights methodological challenges in measuring energy and carbon impacts and calls for interdisciplinary methods and standardized evaluation models to advance practical sustainability in microservice architectures.
Abstract
Microservices are a popular architectural style adopted by the industry when it comes to deploying software that requires scalability, maintainability, and agile development. There is an increasing demand for improving the sustainability of microservice systems in the industry. This rapid review gathers 22 peer-reviewed studies and synthesizes architectural tactics that improve the environmental sustainability of microservices from them. We list 6 tactics that are presented in an actionable way and categorized according to their sustainability aspects and context. The sustainability aspects include energy efficiency, carbon efficiency, and resource efficiency, among which resource efficiency is the most researched one while energy efficiency and carbon efficiency are still in the early stage of study. The context categorization, including serverless platforms, decentralized networks, etc., helps to identify the tactics that we can use in a specific setting. Additionally, we present how the evidence of optimization after adopting these tactics is presented, like the measurement unit and statistical methods, and how experiments are generally set up so that this review is both instructive for our future study and our industrial practitioners' interest. We further study the insufficiencies of the current study and hope to provide insight for other researchers and the industry.