Enhancing Cloud Computing Performance for Fast Online Services Delivery Using a Novel Hybrid Task Scheduling Algorithm

Abstract

Forwarding many tasks to the cloud server simultaneously degrades cloud computing performance in this regard, time-dependent sectors like banks, healthcare, airlines, online shopping, and more suffer from slow delivery of cloud computing services unless a suitable task scheduling technique is implemented. The existing literature-based evidence reveals that effective task scheduling in cloud computing optimizes resource usage, boosts fault tolerance, facilitates scalability, lowers costs, and reduces energy consumption. However, identifying task-scheduling algorithms that fulfill the stated benefits is a hot research topic. To address these concerns, the study explored the type of task scheduling algorithms utilized, the dataset source, the number of tasks given, the number of cloud hosts, the simulation toolkits, the kind of evaluation metrics used, the number of virtual machines, and the scheduling techniques that performed best overall from 36 recent, relevant and reliable research works; applied one of the top performing task scheduling algorithms in most reviewed studies i.e., Genetic Algorithm(GA) along with task scheduling algorithm used in none of the reviewed research works i.e., Earliest Deadline First (EDF) as novel contributions, Compares the results of the proposed approach with another study, which employed the same datasets and evaluation metrics. The overall experiment results show that the proposed approach outperforms GA and GWO algorithms in terms of minimizing Makespan across all workloads and optimizing resource utilization on all workloads, except on workloads 300 and 500, where it performs less than GWO algorithms. GWO algorithm is slightly performing better than the proposed approach on workloads 500, 600, 700, 800, 900, and 1000 in maximizing throughput while performing less than the proposed approach on workloads 100, 200, and 300 in maximizing throughput. Compared to the proposed approach (GA-EDF combination) and GWO algorithm, the GA task scheduling algorithms demonstrate poor performance based on Makespan, average resource utilization, and throughput.