Date of Graduation
Master of Natural and Applied Science in Computer Science
Cloud performance relies on carefully selecting datacenters, load-balancing algorithms, and resource utilization. The datacenter selection policy is crucial in ensuring the choice of an efficient and effective datacenter. Researchers found that using evolutionary algorithms can provide better datacenter selection policies for cloud computing environments. In this thesis, I have started with a genetic algorithm to find the most suitable datacenter for specific userbases. I aimed to improve overall response and data processing times within the cloud environment. The genetic algorithm is a well-known nature-inspired evolutionary algorithm that emulates nature's selection, crossover, and mutation mechanisms across generations to discover optimal problem solutions. Results showed that the genetic algorithm provides better results than other algorithms, yet consistency in the results is missing. For this reason, I explored the swarm intelligence algorithms for the datacenter selection policy. Particle swarm optimization algorithms utilize collective knowledge of the swarm to navigate toward optimal solutions, leveraging insights from neighboring particles. However, the swarm intelligence algorithm's performance was subpar compared to the genetic algorithm. Meanwhile, differential evolution relies on a mutation-driven evolutionary process, enabling comprehensive search space exploration to discover the most suited datacenter. Because of this reason, I worked with the differential evolution algorithm to examine performance in datacenter selection. The outcomes of my thesis unequivocally demonstrate the superiority of the proposed evolutionary algorithms over existing baseline datacenter selection methods, like the closest datacenter selection policy and optimized response time policy. The results also indicate that the evolutionary algorithms can match or outperform the baseline datacenter selection policies when the number of datacenters is reduced. Moreover, the results show significant improvements in response and data processing times for my proposed algorithms compared with existing evolutionary-based datacenter selection policies. Furthermore, in this thesis research, I show that differential evolution consistently delivers better response and data processing times among the three evolutionary algorithms.
Cloud, Userbase, Datacenter, Datacenter Selection Policy, Response Time, Data Processing Time, Evolutionary Algorithms, Genetic Algorithm, Particle Swarm Optimization, Differential Evolution.
© Shusmoy Chowdhury
Chowdhury, Shusmoy, "Optimal Cloud Datacenter Selection Using Evolutionary Algorithms" (2023). MSU Graduate Theses. 3914.
Available for download on Sunday, January 19, 2025