Meta-Heuristics Based Load Balancing Optimization in Cloud Environment on Underflow and Overflow Conditions

Meta-Heuristics Based Load Balancing Optimization in Cloud Environment on Underflow and Overflow Conditions

Amanpreet Kaur (IK Gujral Punjab Technical University, Kapurthala, India), Bikrampal Kaur (Chandigarh Engineering College, Landran, Mohali, India) and Dheerendra Singh (Chandigarh College of Engineering and Technology, Chandigarh, India)
Copyright: © 2018 |Pages: 18
DOI: 10.4018/JITR.2018100110

Abstract

In cloud environment, the main challenge is load balancing as it requires distributing the load among many various virtual machines (VM) while avoiding underflow and overflow conditions. In this article, the question that which load-balancing (overflow or underflow) management will better improves the performance and quality of service has been answered with a number of experiments. For experiment purpose, scientific workflow DAG files has been used with one host configuration. Ant Colony Optimization (ACO) and BAT algorithm are used for checking underflow and overflow conditions respectively for VMs. In proposed work, initially the workflow is parsed by Predict Earliest Finish Time (PEFT) heuristic to generate initial seed for meta-heuristic algorithms which will optimize the VM in terms of makespan and cost of execution. Different workflows have been used with varying number of VMs from 2 to 20. The results shows that makespan analysis is approximately overlapped for different workflow tasks and shows no significant difference however, the cost analysis show a significant change for overflow and underflow identification with cost for overflow condition is reduced significantly.
Article Preview
Top

Introduction

Cloud computing technology is based on heterogeneously distributed computing model for providing hardware, software and platform for managing and developing applications. Cloud computing is further based on virtualization and internet technologies to provide these services to cloud users whenever they demand on basis of pay-as-you-go. Cloud computing has collection of heterogeneous datacenters providing memory, storage, network bandwidth, computational power, and application development software. The responsibility of the cloud service provider is to provide services in seamless manner. In addition, cloud services also involve dynamically provisioning the sharable resources among the user request. The resources are allocated and de-allocated optimally as per the application demand while considering resource availability and performance requirements based on Quality of Service parameters like energy utilization, cost, time, resource utilization and throughput (Liu, 2014).

A dynamic collection of heterogeneous resources is available at cloud datacenter. It keeps the both dynamic and static information regarding these resources. Static information includes available datacenter information about its storage, processor cycles, memory and storage/memory allocated to VMs and throughout the cloud datacenter, this information is constant. Whereas, load allocated to hosts of datacenter is included in the dynamic information of datacenter, also, at a particular instance, the number of threads/ tasks running, running states of tasks, particular task utilizing number of CPU cycles, and tasks status during their execution. It is necessary to update in real time and at regular intervals, this dynamic information, so that the handling of dynamic requests from cloud user can be done in minimum response time.

In this paper, a comprehensive multi-objective Load balancing framework for cloud environment has been proposed, implemented and analysed for underflow and overflow conditions of VMs using workflow’s DAG as input files. It works for both cloud provider and end user perspective.

Cloud Infrastructure as a Service (IAAS) has been referred for virtual resources which are provisioned dynamically through resource pooling.

The requested cloud IAAS resources are allocated to tasks in the form of VMs such that each task is executed on a single VM whereas, a single VM can be allocated to multiple tasks. The cloud resources are provisioned dynamically considering resource heterogeneity and resource pooling. This multi-objective framework optimizes total execution time (makespan) and cost (budget) incurred for resource usage when tasks are mapped onto resources (VMs). From the cloud provider prospect, the proposed framework for load balancing aims to achieve high resource utilization such that the tasks are mapped onto resources without over utilizing or underutilizing any Virtual Machine. Further, the framework is analysed on the basis of makespan and cost metrics to determine whether underutilization or overutilization of VMs must be more concerned which will affect the overall performance of the load balancer. The end user, on the other hand, is ensured high availability of resources and reduced cost.

Complete Article List

Search this Journal:
Reset
Open Access Articles
Volume 14: 4 Issues (2021): Forthcoming, Available for Pre-Order
Volume 13: 4 Issues (2020)
Volume 12: 4 Issues (2019)
Volume 11: 4 Issues (2018)
Volume 10: 4 Issues (2017)
Volume 9: 4 Issues (2016)
Volume 8: 4 Issues (2015)
Volume 7: 4 Issues (2014)
Volume 6: 4 Issues (2013)
Volume 5: 4 Issues (2012)
Volume 4: 4 Issues (2011)
Volume 3: 4 Issues (2010)
Volume 2: 4 Issues (2009)
Volume 1: 4 Issues (2008)
View Complete Journal Contents Listing