Aburukba, RaafatKhan, Mueez Ahmad2025-01-212025-01-212024-1235.232-2024.46https://hdl.handle.net/11073/25779A Master of Science thesis in Computer Engineering by Mueez Ahmad Khan entitled, “Optimizing Energy Consumption In Cloud Datacenters”, submitted in December 2024. Thesis advisor is Dr. Raafat Aburukba. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).Cloud computing has become a cornerstone of modern technology, enabling scalable and efficient resource utilization. However, the rapid growth in demand has resulted in significant energy consumption, posing challenges for sustainability and operational efficiency in cloud datacenters. This thesis addresses the critical issue of energy optimization in task scheduling within cloud datacenters, where increasing demand has led to significant energy consumption and environmental impact. Tasks with varying complexities are allocated to cores with unique specifications, aiming to minimize energy usage while maintaining operational efficiency. A comprehensive mathematical model is proposed to minimize energy consumption when assigning tasks to cores in a datacenter. The model is validated using exact solutions methods for small-scale instances. To further test the model on large scale problems, two hybrid heuristic algorithms based on Genetic Algorithm (HGA) and Simulated Annealing (HSA), are proposed. Parameter tuning for HGA and HSA was performed to further improve the solution quality and reduce execution time. Experiments were conducted on small, medium, large and x-large problem sets to test the scalability of the heuristics. Small size problem set was used to compare the heuristic quality to the exact solutions which showed that the heuristics provide a higher energy consumption by around 5% compared to the exact solution but with approximately 50% faster execution time. This proves that both heuristics provide a near optimal solution when compared to the exact solution with a much faster execution time. For medium-sized problems, HGA provided a lower energy consumption of around 6% over HSA, with an approximately 35% longer execution time. In large and extra-large problems, HGA outperformed HSA in providing a lower energy consumption by around 10%, but required around 23% more time for execution. This demonstrates that HSA is more suitable for scenarios where quick convergence is prioritized, whereas HGA is better suited for applications that require minimum energy consumption.en-USEnergy OptimizationTask AllocationCloud DatacentersGenetic Algorithm (GA)Simulated Annealing (SA)Thesis