Smart Resource Optimization for Cloud Infrastructure Using Machine Learning

Govindaraju, Hemadhri (2025) Smart Resource Optimization for Cloud Infrastructure Using Machine Learning. Masters thesis, Dublin, National College of Ireland.

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

Abstract

The increasing demands of scalable, cost-efficient and intelligent cloud infrastructure has revealed the weakness of conventional reactive auto-scaling patterns adopted in the systems such as AWS. These threshold-based methods tend to be oversupplied or slow in responding to dynamic workloads. This study presents a machine learning (ML)-based predictive auto-scaling topology, which uses the Long Short-Term Memory (LSTM), Prophet, and a Hybrid model to predict CPU utilization and implement EC2 instance automatic scaling. The models can predict the future CPU load and take scaling decisions using real-time system metrics identified using CloudWatch. They were evaluated by running live EC2 instances and logging predictions, which were then visualized in Grafana. Findings reveal LSTM has high ability to predict, particularly in sharp variation, whereas Prophet renders a smooth trend, but not very fast in responding to spikes. The Hybrid model that has combined 70% LSTM and 30% Prophet output has the best overall performance that not only reduced false positives, but also improved SLA compliance, and optimized cost. The ML models greatly enhanced the CPU utilization and minimized unnecessary scaling activities and operations in comparison to the legacy system used by AWS of Auto Scaling Group. According to this study, predictive scaling, complemented by hybrid learning strategies and real-time observability, can be a more adaptive and efficient alternative to the cloud resources management.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Heeney, Sean UNSPECIFIED
Uncontrolled Keywords:	Auto-scaling; Machine Learning; LSTM; Prophet; Predictive Scaling; EC2; CloudWatch; Grafana
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:	31 Aug 2026 14:46
Last Modified:	31 Aug 2026 14:46
URI:	https://norma.ncirl.ie/id/eprint/9706

Actions (login required)

View Item