Pijnacker, Bjorn (2024) Estimating Container-level Power Usage in Kubernetes. Master's Thesis / Essay, Computing Science.
|
Text
mCS2024PijnackerB.pdf Download (1MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
Toestemming.pdf Restricted to Registered users only Download (183kB) |
Abstract
Energy efficiency in cloud computing, and specifically Kubernetes, has been a major topic of research in the past years as cloud datacenters grow and the importance of minimizing carbon output increases. Much of this research focuses on the total energy usage of a Kubernetes cluster and attempts to optimize this by various methods such as energy-aware scheduling and datacenter- or cluster-wide metrics, without regard for individual workloads. A lesser explored aspect which can provide useful insights into Kubernetes power usage is a measure of power usage at the Kubernetes container level; thereby providing insight into the power usage for each workload in a cluster. In this research project, we experimentally evaluate the state-of-the-art Kubernetes power estimation tooling: the Cloud-Native Computing Foundation's Kepler. This tool is evaluated on datacenter-grade hardware where its total Kubernetes node measurements and its container power attribution for each of the available configurations and available external sources is considered. We find that this tool does not produce satisfactory power usage metrics in regard to container power attribution and that there are significant limitations in several of the available node power measurement strategies. To combat the limitations in Kepler, we create our own tool named KubeWatt based on a recently introduced power mapping model. The architecture and implementation of KubeWatt are discussed along with an experimental validation of its features and measurements. During this work, several limitations as well as some interesting findings are discussed which may provide avenues for future research.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Andrikopoulos, V. and Setz, B. |
| Degree programme: | Computing Science |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 21 Nov 2024 07:48 |
| Last Modified: | 21 Nov 2024 07:48 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/34420 |
Actions (login required)
 |
View Item |