An Optimal Bio Inspired Genetic Algorithm for Load Balancing for Cloud Data Centre

Author Name(s): P.S.Latha Kalyampudi, P.Venkata Krishna
Author Email: pslathakalyampudi@gmail.com

Abstract

Demand for the high performance computing for industry, education and research is continuously motivating the application development industry to manoeuvre the existing and the upcoming applications towards the cloud. The deployment or the migration of the new or existing application on cloud data centres demands greater skills for development and maintenance of the applications. The biggest challenges for the data centre service providers are to balance the bottleneck of performance and cost. Many application or service owners have demanded for higher performance at a higher cost. Nevertheless, the traditional customers base have attracted towards cloud computing due to the low cost and higher performances. Thus for the data centre service providers it is the challenge to make the highest performance available towards the customers in the least cost. In order to make this challenge possible, the data centre service providers deploy various load balancing strategies for the most effective use. A number of research attempts are made towards achieving the best possible load balancing strategy by number of parallel research attempts. Many parallel research attempts have demonstrated the adaptive, strategic and just in time scheduling and load balancing algorithms with notable reduction in time for scheduling. Nonetheless, the previous works are always outperformed by the new algorithms proposed by other research works. In the recent time, the use of genetic algorithms and genetic optimization algorithms has demonstrated higher performances. However, the uses of biogenetic algorithms have the chance to improve the performance further as always. Thus, this work proposes an additional optimization using proposed bio genetic optimization for effective balancing the data centre tasks. The proposed method demonstrates the 4% higher performance compared to the existing methods for a less loaded data centre and 16% improvements for a highly loaded data centre.

Introduction

The rapid growth in the cloud computing domain is push forwarded by the industry to match the customer demands and propelled by the research outcomes. The application owners had constant demand for the high and scalable performance measures, which will not inflate the cost of the application deployments and maintenance. The paradigm of cloud computing fulfils these requirements from the client, data centre service providers and the application or service owners. The notable work by A. Weiss et al. [1] has proven the benefits of the cloud computing and defined it as the future of computing and services. The advantage from cloud computing is majorly achieved by the virtualization technique. The virtualization is a widely accepted method and demonstrated by various parallel research outcomes. The notable outcomes are by P. Barham et al. [2] on Xen as the VMM and by the J. Fisher et al. [3] on the hardware support. The basic principle of virtualization is the optimal utilization of the resources as proven by C. Fangzhe et al. [4]. Nevertheless, the allocation of the tasks on the virtual machine cannot be done manually, hence the multiple task allocation and assignment algorithms have evolved. The notable works by F. Chang et al. [5] and by J. Ren et al. [6] are a must to be considered for identifying the initial challenges of task allocation on cloud data centres. The complete automation of the task, called task scheduler and the optimization on cloud data centre is first introduced by H. Qiyi et al. [7]. Thereafter a number of research attempts are made to further optimize the task scheduling on cloud. It is identified by multiple experiments that the optimization performance can be higher with the inclusion of biogenetic dynamic programming methods as Ant-Colony-Optimization or ACO. This method was first introduced by M. Dorigo et al. [8] in the year of 2005 and the further improvements to make it applicable for cloud computing was produced by M. Birattari et al. [9] in the year of 2006. As stated by L.M. Gambardella et al. [10], the ACO is the best random search algorithm till date.

Conclusion

The modern application hosting and development industry demands higher performance and availability. The higher performance always cannot be reached with higher infrastructure. Thus the research demands better algorithms for load allocation and balancing. A number of research attempts have tried the optimization on load balancing and the recent trends demonstrated the use of genetic optimization techniques. Nevertheless, this work identified the shortcomings and proposed a new method for optimization using parametric Genetic-Colony-Optimization. The outcomes from the proposed algorithm have demonstrated a higher response time and more normal load distribution than the present methods. Thus, this work contributes towards an improvement and betterment of the load balancing strategies for better computing on cloud.

 

871 total views, no views today

Download PDF File

About the author: admin