Optimizing Container Resource Allocation by Right Sizing using Historical Timeseries Metrics in ARIMA Model

Kuriyakose Kadavil, Libin Tom (2023) Optimizing Container Resource Allocation by Right Sizing using Historical Timeseries Metrics in ARIMA Model. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration manual) Download (747kB) | Preview

Abstract

Containers have revolutionized the deployment and management of applications in cloud computing environments, offering significant advantages such as lightweight and portable packaging, faster startup times, and reduced resource overhead. While containers excel in resource efficiency, the lack of resource limitations can lead to overconsumption or underutilization, especially in resource-constrained edge cloud environments. In edge cloud environments, which are located closer to end-users and devices, computing resources are inherently limited due to factors like physical space, power availability, and constrained network bandwidth. Optimizing resource usage in edge cloud environments is crucial to ensure smooth operation, meet quality of service requirements, and maximize performance. Right sizing containers plays a vital role in achieving resource optimization by allocating the appropriate amount of compute resources based on application requirements. Right sizing offers advantages such as cost optimization, improved performance, scalability, stability, reliability, and simplified resource management. While container orchestrators like Kubernetes provide features like Vertical Pod Autoscaler (VPA) for resource allocation optimization, relying on historical resource usage metrics for right sizing containers offers several advantages over real-time metrics. Historical metrics provide a more comprehensive and detailed view of resource utilization patterns, enabling accurate recommendations for resource requests and limits. They also allow for the identification of trends and seasonal variations, facilitating informed decision making for resource allocation. This research aims to explore approaches for right sizing container resource allocation by analyzing real historical resource usage metrics of a private company. By implementing Time Series forecasting model ARIMA (Auto Regressive Integrated Moving Average) to predicted the future CPU requirements. Findings show a potential cost reduction of up to 60% over a year compared to traditional fixed resource limits. While acknowledging the rough nature of the estimation, the study underscores the potential benefits of the approach.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Lugones, Diego UNSPECIFIED
Uncontrolled Keywords:	Container Orchestration; Resource Management; Time Series Analysis; Predictive Modeling; Auto Regressive Integrated Moving Average (ARIMA); Kubernetes
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Tamara Malone
Date Deposited:	09 Oct 2024 17:42
Last Modified:	09 Oct 2024 17:42
URI:	https://norma.ncirl.ie/id/eprint/7087

Actions (login required)

View Item