Abstract:

Abstract Cloud computing has emerged as a crucial platform for managing and executing time-constrained scientific applications, typically represented by workflow models and their scheduling. The scheduling of workflow applications in cloud computing poses a significant challenge, as they consist of numerous tasks with complex structures involving processing, data entry, storage access, and software functions. To address this challenge, users are provided with a convenient and cost-effective approach to run workflows on rented on-cloud Virtual Machines (VMs) at any time and from anywhere. With the growing dominance of pay-as-you-go pricing models in cloud services, extensive research has been conducted to minimize the cost of workflow execution by developing customized VM allocation mechanisms. However, most existing approaches assume static task execution times in the cloud, which can be estimated in advance. Unfortunately, this assumption is highly impractical in real-world scenarios due to performance variations among VMs. In this study, we propose a custom workflow scheduling algorithm designed to handle deadline-constrained workflows with random arrivals and uncertain task execution times, while ensuring higher CPU utilization. Our algorithm supports the use of containers to manage targets and optimize resource utilization, thereby reducing the overall cost of infrastructure resources and meeting individual workflow deadline constraints. Simulation results demonstrate that the proposed algorithm outperforms existing approaches in terms of rental costs and resource utilization efficiency.

Keywords: