Heuristic Load Balancing Based Zero Imbalance Mechanism in Cloud Computing

• Published in: Journal of grid computing Vol. 18; no. 1; pp. 123 - 148
• Main Authors: Kong, Lingfu, Mapetu, Jean Pepe Buanga, Chen, Zhen
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.03.2020; Springer Nature B.V
• Subjects: Algorithms; Cloud computing; Completion time; Computer networks; Computer Science; Distributed processing; Heuristic; Heuristic methods; Load balancing; Management of Computing and Information Systems; Optimization; Priority scheduling; Processor Architectures; Random access memory; Resource utilization; Schedules; Task scheduling; User Interfaces and Human Computer Interaction; User requirements; User satisfaction; Virtual environments; Task scheduling; Cloud computing; Heuristic algorithms; Load balancing; Zero imbalance
• ISSN: 1570-7873; 1572-9184
• DOI: 10.1007/s10723-019-09486-y

Cloud computing using virtualization technology has emerged as a new paradigm of large-scale distributed computing. One of its fundamental challenges is to schedule a vast amount of heterogeneous tasks while maintaining load balancing amongst different heterogeneous virtual machines (VMs) to meet both cloud users and providers’ requirements, such as minimum makespan low monetary costs, and high resource utilization. This problem is often classified as, NP-hard optimization, and while many heuristic algorithms have attempted to solve the NP-problem. However, they fail in load balancing and lower running times when the number of tasks grows exponentially, while that of VMs with set of resources, such as CPU, memory RAM and bandwidth remains stagnant. To solve the NP-problem effectively, we propose a fast heuristic algorithm based on the zero imbalance approach, as a new concept in the heterogeneous environment. Specifically, this approach focuses on minimizing the completion time difference among heterogeneous VMs without priority methods and complex scheduling decision which often subject the heuristic algorithms to the particular cloud configuration. The proposed approach defines two constraints, optimal completion time and earliest finish time which take account the task transfer time onto network bandwidth of VM to achieve load balancing and task scheduling effectively. The experimental results below show that the proposed algorithm effectively solves the NP-hard optimization problem better than existing heuristic algorithms by satisfying cloud users and providers’ requirements.