Leveraging K-means Clustering for Multi-objective Task Scheduling in Cloud-based Scientific Workflows
July Lwin, Tin Zar Thaw · 2025
In the cloud environment, efficient task scheduling for scientific workflows presents significant challenges, especially when addressing multiple conflicting objectives such as minimizing execution time and cost. This study proposes a novel multiobjective task scheduling approach utilizing the K-means clustering algorithm, tailored for optimizing the Montage, and Sipht, well-known scientific workflows. K-means enhances resource allocation by grouping tasks with similar characteristics, enabling better load balancing and scalability. Experimental results demonstrate notable improvements in makespan and total execution time compared to traditional scheduling methods. This work underscores the potential of the proposed approach to advance task scheduling techniques for scientific workflows, paving the way for broader applications in the cloud environment.