HDRLGA: Hybrid deep reinforcement learning and genetic algorithm task scheduling approach in cloud computing.

Abstract

Task scheduling in cloud computing is a critical challenge that directly impacts the performance and cost-effectiveness of cloud environments. This paper presents a hybrid task scheduling algorithm combining Deep Reinforcement Learning (DRL) and Genetic Algorithm (GA) to achieve multiple optimization objectives including minimizing makespan, reducing overall cost, average turnaround and degree of imbalance. HDRLGA method leverages DRL to predict optimal task-to-resource mappings dynamically, while GA fine-tunes the schedule by exploring a wide range of possible configurations. Experimental results demonstrate that the hybrid approach outperforms traditional scheduling methods shortest job first, largest job first, first come first serve, max-min and min-min in terms of makespan and degree of imbalance. The algorithm’s effectiveness is validated through simulations using various cloud scenarios, highlighting its potential for real-world cloud environments.