Indian Journal of Science and Technology

Abstract

Background/Objectives: Load balancing algorithms are a type of algorithm that assists with efficient workloads spread. To reduce server overload, increase resource usage, lower latency, and maximize throughput, efficient distribution is required. There are several load balancing algorithms, all of these algorithms have flaws that render them ineffectual in many real-world settings. The current paper aims to improvise the algorithm by proposing a heuristic to decide the next job to be allotted to reduce the net response time. This is achieved by storing the incoming jobs in a priority queue (heap) and using ageing to prevent starvation. Methods: In the current research, we assign the incoming client request to the virtual machine such that estimated finish time of the service is minimum. We calculate finishTime for all the virtual machines, then the VM which gives the minimum finishTime is assigned the incoming client request. Once a client request is assigned to a VM, it gets added to the virtual Machine Queue. The virtual Machine Queue behaves like a priority queue; it gives priority to the task which has the least instruction Count. This ensures that the Net Response Time of the load balancer is minimized. Findings: Significant improvements have been achieved with respect to response time and throughput. We calculated the Response Time and the Throughput Time for the Load Balancing Using Estimated Finish Time Algorithm compared to IEFTA. The numbers of client requests were varied by a factor of 10, starting from 10, and we got an efficient distribution. Novelty: By using our proposed Algorithm Pseudocode of Load Balancing Algorithm, different coefficient of VMs were calculated. We compared the Estimated Finish Time Algorithm to IEFTA, we suggest an approach that addresses this limitation while also improving responsiveness and processing time.

Keywords: Cloud Computing; Distributed Systems; Load Balancing; Virtual Machines; Data Migration