Hybrid Genetic Algorithm-Based QoS-Aware Task Scheduling In Cloud Environments

Abstract

Enhancing multiple Quality of Service (QoS) parameters simultaneously remains a significant challenge in cloud computing (CC), particularly when the delivered QoS fails to meet end-user expectations. This study introduces an improved Smart Message Passing Interface Approach (SMPIA) integrated with two optimization techniques: Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). The proposed hybrid models, namely GA-SMPIA and PSO-SMPIA, are designed for efficient task scheduling and resource allocation in cloud environments. The primary objective is to minimize makespan and total execution time while maximizing resource utilization. A key contribution of this research is the formulation of a novel multi-objective cost function that determines the maximum cost associated with each transaction flow—an aspect not sufficiently addressed in earlier studies. This function incorporates multiple parameters, including flow load, makespan load, virtual machine (VM) capacity, and execution speed.Additionally, telecommunication transactions are categorized based on flow types, and an allocation matrix is constructed to map transaction flows to appropriate VMs. This matrix guides the routing of transactions to optimize performance. The study also evaluates how makespan and execution time influence overall resource utilization. Experimental results demonstrate that PSO-SMPIA achieves superior performance in terms of average resource utilization compared to existing methods such as SMPIA, Fuzzy SMPIA (FSMPIA), Optimized SMPIA (O-SMPIA), and GA-SMPIA. However, FSMPIA and O-SMPIA show better results in minimizing makespan and execution time individually. Notably, GA-SMPIA provides a balanced improvement across all key metrics by effectively reducing execution time, minimizing makespan, and enhancing resource utilization.

Overall, the proposed approaches significantly improve QoS delivery in cloud computing environments, making them suitable for real-world applications requiring efficient and reliable resource management.