aalto1 untyped-item.component.html
Analyzing the energy impact of caching strategies in containerized cloud applications
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Science |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
Department
Major/Subject
Mcode
Degree programme
Language
en
Pages
79
Series
Abstract
The increasing energy demand of cloud-native applications, especially those that depend on large language models and intensive backend services, has generated a need to have sustainable modern software applications. While infrastructure-level optimizations are being implemented and explored, software-level decisions, as changing strategies, remain understudied. This thesis researches how different caching strategies and cache hit ratios affect the energetic consumption of containerized applications deployed in Kubernetes environments. An experimental environment was designed implementing PostgreSQL, Redis, In-Memory, and Hybrid caching strategies. Energy consumption was estimated using Kepler, a Kubernetes-native monitoring tool, across two scenarios: application-level (backend-only) and system-level (all pods in the request flow). Results show that increasing the cache hit ratio leads to an almost linear decrease in energy consumption per request. All the caching strategies surpassed in efficiency the baseline without caching, with In-Memory Cache being the most energy-efficient, and Hybrid Cache offering the best balance between energy and latency. Linear regression models built for each strategy demonstrated strong predictive power (adjusted 𝑅2 > 0.99). The findings demonstrate that caching usage at the software level significantly reduces the energy consumption of containerized applications and that its efficiency is an important factor to achieve a sustainable software design. The study contributes practical insights and predictive models to inform the development of more sustainable backend systems.
Description
Supervisor
Hellas, ArtoThesis advisor
Ebrahimy, RazgarHellas, Arto