Intelligent Multi-Objective Decision Support System for Efficient Resource Allocation in Cloud Computing

Abstract

The dynamic allocation of materials within cloud systems is essential for optimizing system architecture, enhancing energy efficiency, and ensuring compliance with Service Level Agreements (SLA). to address workload imbalance and resource overload issues, this research introduces an Intelligent Multi-Objective Decision Support System (IMODSS) for resource allocation in cloud systems. The proposed framework leverages the novel integration of the Modified Feeding Birds Algorithm (ModAFBA) with the Deep Reinforcement Learning (DRL)-based Q-Learning algorithm to enhance the adaptability and effectiveness of resource management. By combining the dynamic clustering abilities of ModAFBA with the adaptive decision-making of Q-learning, IMODSS effectively prioritises tasks, balances workloads throughout the virtual machine (VM), and improves energy efficiency. Experimental validation using Python and CloudSim demonstrates that IMODSS notably outperforms traditional methods. Specifically, the proposed system reduces makespan by 15% to 20%, energy consumption by 18% to 22%, and VM migrations by 20% to 25% compared to existing cloud-based resource allocation models of HBCA, MOPSO, and TPOSIS. Also, the integration of Q-Learning strengthens the system to manage QoS parameters, such as CPU and memory utilization efficiency and SLA violation control. Therefore, the IMODSS framework effectively scales under varying workload conditions and is a promising solution for next-generation cloud computing environments.