TITLE:
Efficient Resource Allocation in Cloud IaaS: A Multi-Objective Strategy for Minimizing Workflow Makespan and Cloud Resource Costs
AUTHORS:
Jean Edgard Gnimassoun, Dagou Dangui Augustin Sylvain Legrand Koffi, Akanza Konan Ricky N’dri
KEYWORDS:
Cloud Infrastructure, Multi-Objective Scheduling, Resource Cost Optimization, Resource Utilization, Scientific Workflows
JOURNAL NAME:
Open Journal of Applied Sciences,
Vol.15 No.1,
January
26,
2025
ABSTRACT: The ease of accessing a virtually unlimited pool of resources makes Infrastructure as a Service (IaaS) clouds an ideal platform for running data-intensive workflow applications comprising hundreds of computational tasks. However, executing scientific workflows in IaaS cloud environments poses significant challenges due to conflicting objectives, such as minimizing execution time (makespan) and reducing resource utilization costs. This study responds to the increasing need for efficient and adaptable optimization solutions in dynamic and complex environments, which are critical for meeting the evolving demands of modern users and applications. This study presents an innovative multi-objective approach for scheduling scientific workflows in IaaS cloud environments. The proposed algorithm, MOS-MWMC, aims to minimize total execution time (makespan) and resource utilization costs by leveraging key features of virtual machine instances, such as a high number of cores and fast local SSD storage. By integrating realistic simulations based on the WRENCH framework, the method effectively dimensions the cloud infrastructure and optimizes resource usage. Experimental results highlight the superiority of MOS-MWMC compared to benchmark algorithms HEFT and Max-Min. The Pareto fronts obtained for the CyberShake, Epigenomics, and Montage workflows demonstrate closer proximity to the optimal front, confirming the algorithm’s ability to balance conflicting objectives. This study contributes to optimizing scientific workflows in complex environments by providing solutions tailored to specific user needs while minimizing costs and execution times.