Evaluating Scheduling Efficiency for Complex Scientific Workflows in Cloud Infrastructures

Abstract

Scientific workflow scheduling in Cloud computing is critical for efficiently managing data-exhaustive and compute-rigorous applications. With the rising demand of distributed computing, the interest in feasible assignment planning calculations has developed primarily. Revised text: This paper presents a comprehensive survey of advanced scheduling techniques focusing on minimizing energy consumption and improving resource utilization. Various scheduling algorithms, including heuristic-based, meta-heuristic-based, and reinforcement learning-based methods, are analyzed and compared. Additionally, the paper addresses the challenges of scheduling workflows with complex dependencies, offering a novel multi-objective workflow scheduling algorithm using reinforcement learning. The algorithm outperforms current methods in terms of makespan and energy consumption. Finally, the paper highlights open research issues and future directions in scientific workflow scheduling for distributed computing. Through our experiments, we have achieved significant improvements in scheduling efficiency, demonstrated by various performance metrics illustrated in our graphs.