EWPO-IA3C: An Optimized Deep Reinforcement Learning Framework for Task Scheduling in Multi-Cloud Environments

Abstract

The Multi-Cloud Environment (MCE) is a widely adopted computing model for providing on-demand services. Task scheduling in such environments is a critical issue, and numerous algorithms have been developed to address it, with most mainly focusing on reducing the overall completion time, often neglecting the optimization of total service cost. Additionally, many of these methods are designed for single-cloud rather than multi-cloud settings. To tackle these problems, this paper proposes a framework named Exponential Wave Propagation Optimization-enabled Improved Asynchronous Advantage Actor Critic (EWPO-IA3C) algorithm for task scheduling in MCEs. The proposed EWPO-IA3C is evaluated in terms of makespan, cost, and execution time, with respective results of 4088.12 ms, 40.881$, and 4.088s.