Understanding the impact of optimised resource allocation in cloud computing for improved cost and performance

Shatoji, Deepa Ashok (2025) Understanding the impact of optimised resource allocation in cloud computing for improved cost and performance. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (864kB) | Preview

Abstract

The cloud computing environments are under increasing strain to perform high performance and cost effectiveness as well as handle multifarious and changing workload on a multifarious group of providers. Most of the rule-based allocation strategies fail to scale to the heterogeneous infrastructures resulting to over provisioning, under-utilisation and performance bottlenecks. This paper explores how machine-learning-based optimisation can be used to optimize resource allocation in multi-cloud systems. An empirical data set of compute, storage, network and latency metrics was employed to design key performance indicators, such as: the Resource Efficiency Score, the Cost Proxy, the Performance Ratio and the Energy Proxy that allow a multidimensional analysis of cloud workload behaviour. Four models of supervised learning were trained to predict resource efficiency levels by using four supervised learning models- Logistic Regression, Random Forest, XGBoost and Gradient Boosting. The best performance was offered by the Logistic Regression, which had an accuracy of 96 percent and F1-score of 95.96, which is better than the ensemble methods because the features are well separated, and the classes are well balanced. KPI revealed moderate utilisation and high latency and response times, which shows a chance to optimise highly distributed multi-cloud deployments. The findings show that machine-learning-based solutions can greatly enhance the reliability of predictions and assist in proactive scaling decisions as well as decrease cost-performance trade-offs. The work adds a feature-engineering framework of structure, a resource-efficiency classifier that has been verified, and implementation insights of useful information on real-world smart, cost-effective optimisation techniques in the current cloud systems.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Siddig, Abubakr UNSPECIFIED
Uncontrolled Keywords:	Cloud Computing; Resource Allocation; Machine Learning; Cost Optimisation; Performance Enhancement
Subjects:	T Technology > T Technology (General) > Information Technology > Cloud computing Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	01 Sep 2026 11:00
Last Modified:	01 Sep 2026 11:00
URI:	https://norma.ncirl.ie/id/eprint/9736

Actions (login required)

View Item