Title: A hybrid algorithm for workflow scheduling in cloud environment

Authors: Tingting Dong; Li Zhou; Lei Chen; Yanxing Song; Hengliang Tang; Huilin Qin

Addresses: School of Information, Beijing Wuzi University, Beijing, 101149, China; Faculty of Information Technology, Beijing University of Technology, Beijing, 100000, China ' School of Information, Beijing Wuzi University, Beijing, 101149, China ' School of Information, Beijing Wuzi University, Beijing, 101149, China ' School of Information, Beijing Wuzi University, Beijing, 101149, China ' School of Information, Beijing Wuzi University, Beijing, 101149, China ' School of Information, Beijing Wuzi University, Beijing, 101149, China

Abstract: The advances in cloud computing promote the problem in processing speed. Computing resources in cloud play a vital role in solving user demands, which can be regarded as workflows. Efficient workflow scheduling is a challenge in reducing the task execution time and cost. In recent years, deep reinforcement learning algorithm has been used to solve various combinatorial optimisation problems. However, the trained models often have volatility and can not be applied in real situation. In addition, evolutionary algorithm with a complete framework is a popular method to tackle the scheduling problem. But, it has a poor convergence speed. In this paper, we propose a hybrid algorithm to address the workflow scheduling problem, which combines deep reinforcement algorithm and evolutionary algorithm. The solutions generated by deep reinforcement learning are the initial population in the evolutionary algorithm. Results show that the proposed algorithm is effective.

Keywords: deep reinforcement learning; evolutionary algorithm; workflow scheduling; multi-objective optimisation.

DOI: 10.1504/IJBIC.2023.130040

International Journal of Bio-Inspired Computation, 2023 Vol.21 No.1, pp.48 - 56

Received: 16 Oct 2021
Received in revised form: 10 Mar 2022
Accepted: 11 Mar 2022
Published online: 04 Apr 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article