Hybrid profitability-aware PSO model with dissimilarity index for cost-efficient and balanced cloud resource allocation

An effective task scheduling of cloud computing systems would involve the trade-off between user-based and performance objectives of users and providers, given the heterogeneous and dynamic workloads. Current scheduling strategies are usually focused on maximizing one goal or they are absent in workload heterogeneity reduction thus, resulting in inefficient use of resources and low profitability. This paper aims to solve these constraints and suggests a Profitability-Aware model using Particle Swarm Optimization with Dissimilarity Index-based Refinement (PSO_DI) architecture to be used in multi-data center cloud-based environment. The two-stage optimization plan is used in the proposed framework. The first stage involves a profitability-conscious, PSO optimizer that allocates tasks to groups of users in batches by jointly maximizing user satisfaction and profit by providers using dynamically adjusted control parameters and profitability-based fitness scores. The second step implements the novel dissimilarity index (DI)-based refinement mechanism which is used to measure the inter-group workload heterogeneity and perform an iterative refinement process of the initial allocation through profit-preservation task reallocation. This integration helps to enhance improved load balance without reducing economic efficiency. The usefulness of the presented PSO-DI framework is assessed on CloudSim with Google Cluster Trace 2019 workloads using a Python framework. The results of the simulation show that PSO_DI attains a higher fitness of approximately 12–18%, lower execution cost of 10–15% and workload dissimilarity of 25–35% when compared to the other approaches of Random allocation, Standard PSO and Hybrid PSO. Furthermore, the proposed method can enhance provider profit by 12–20% and the general profitability balance score by 15–22% and has more rapid convergence behavior. The findings verify that PSO_DI is a powerful, scalable, and cost effective scheduling tool in heterogeneous cloud resources and would be applicable to cloud resources management in practice.