Task Scheduling in Cloud Datacenters: Comparison of Heuristic, Metaheuristic, and Machine Learning Approaches

Yadav, Amit Kumar Chhabiraj (2025) Task Scheduling in Cloud Datacenters: Comparison of Heuristic, Metaheuristic, and Machine Learning Approaches. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (3MB) | Preview
Preview	PDF (Configuration Manual) Download (3MB) | Preview

Abstract

Cloud datacenters are under pressure to reduce energy consumption and maintain high performance. As a result, task scheduling remains a major challenge for scheduling algorithms such as traditional heuristic algorithms, which struggle to handle workload diversity, task to VM allocation, and bursty task arrival patterns. Although heuristic, metaheuristic, and machine-learning schedulers have been studied individually, few works have compared all three paradigms under a common experimental setup. But this project addresses this gap by evaluating seven algorithms: heuristic (Round Robin, Min-Min, Max-Min), metaheuristic ( GA, PSO, ACO ), and a Deep Q-Network (DQN) within a CloudSim Plus environment, which work across six realistic workload scenarios. A total of 210 simulations were run to record energy consumption, makespan, latency, and throughput for all scenarios, and their results show DQN achieves the lowest mean energy usage in five scenarios (496.67 Wh) and outperforms Min-Min by 26% and Round Robin by 49%, while ACO performs best under highly bursty workloads. These findings demonstrate that reinforcement learning is a strong, energy efficient scheduling approach, though its sensitivity to extreme workload variability remains a limitation. In practice, up to 6.3 kWh/day of such RL-based optimization might be saved in a mid-sized datacenter, assuming that there are significant benefits to operations and the environment.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Kazmi, Aqeel UNSPECIFIED
Subjects:	T Technology > T Technology (General) > Information Technology > Cloud computing H Social Sciences > HC Economic History and Conditions > Natural resources > Power resources > Energy consumption 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 11:57
Last Modified:	01 Sep 2026 11:57
URI:	https://norma.ncirl.ie/id/eprint/9746

Actions (login required)

View Item