Javascript must be enabled for the correct page display

Carbon Footprint Monitoring up to Container-Level in Virtualized Environments: A Hardware and Hypervisor-Free Approach

Pol, Ties (2024) Carbon Footprint Monitoring up to Container-Level in Virtualized Environments: A Hardware and Hypervisor-Free Approach. Master's Thesis / Essay, Computing Science.


Download (1MB) | Preview
[img] Text
Restricted to Registered users only

Download (129kB)


The popularity of cloud computing and the increasing demand for computing resources have led to a significant increase in energy consumption. In 2015, the electricity consumed by data centers accounted for 0.9% of global energy usage. This is expected to reach 4.5% in 2025. This increase in demand also drives the carbon footprint of computing environments. This thesis starts by conducting a comprehensive literature review, focusing on methodologies for estimating and mitigating the carbon footprint in computing infrastructures, spanning various models and monitoring tools. Subsequently, this thesis presents an exploratory approach to monitoring carbon emissions down to the level of individual containers. Unlike conventional methods that require hardware, hypervisor, or software access, we try to operate independently of such dependencies. We instead rely exclusively on virtual machine metrics to estimate energy consumption and carbon emissions at the bare metal server level. This result is used to estimate the carbon footprint of the virtual machines and containers in a Kubernetes environment. The validation of our approach involves assessments to ensure that the calculations align consistently across various levels of abstraction. In addition, we align the energy consumption with the number of requests per second by adding load to a container. Although the results of inaccessible levels of abstraction approximately align with the load that is added to the system, there is no internal consistency. The results could be useful to get a rough understanding of the carbon footprint. However, additional research is required to explore whether there exists a consistent method for estimating energy consumption in the absence of hardware or hypervisor-level metrics, and thereafter assess the accuracy of the results.

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: 12 Apr 2024 12:29
Last Modified: 12 Apr 2024 12:29

Actions (login required)

View Item View Item