Digital Twin-Enabled Resource Allocation and Performance Optimisation in Cloud-Based Complex Data Workflows: A Particle Swarm Optimisation Approach
Year: 2026
Type: article
Abstract: In large-scale cloud-based big data systems, optimal resource allocation remains challenging due to the variety of workloads and limited computational resources. Using digital twin technology, we demonstrate our methodology through an optimisation approach to resource allocation and performance tuning for a cloud-based, complex data processing application. Contrary to the conventional heuristic or genetic algorithms popular in the literature, this paper applies Particle Swarm Optimisation (PSO) to efficiently solve the task-to-resource assignment problem. We obtain a synthetic yet realistic workload that includes task execution times, resource usage profiles, and data locality characteristics. The model above addresses the issue by minimising the makespan while imposing penalty constraints for infeasible scheduling and minimising CPU and memory utilisation. Experimental results show that the proposed PSO solution achieves better makespan reduction and improved resource utilisation than the baseline greedy allocation method across various multi-VM configurations. The proposed approach offers a novel and valuable optimisation tool that can be widely applied to cloud workload management by integrating digital twin frameworks.
Author: Ümit Demirbaga
Institution: Bartin University
Language: Turkish
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7160500393