Research On A Clustering-based Deployment Approach For Load Balancing In Cloud Computing Environment

Along with the high-speed development of Internet technology, cloud computing hasbeen paid more and more attention and has been widely used. From the proposal of cloudcomputing to the realization of providing services for enterprises and individuals, cloudcomputing had went through a long course of development. As a new computing model andbusiness model, cloud computing, a kind of new service model is the most promising andvaluable research direction which follows the utility computing, grid computing, distributedcomputing and so on. Cloud computing provides infrastructure, platforms and softwareservice for users. Besides, it provides service for users according to their demands through theInternet. Infrastructure as a service (IaaS) is the basis of cloud computing. IaaS can integratethe computing resources of the cloud data center into a pool of resources throughvirtualization technology. And it allocates on the basis of task specifications and resourcerequests which are submitted by users. IaaS provides users with physical or virtual resourcessuch as calculating, storage and network whose scale can be elastic. Cloud data center deployslots of physical hosts to provide services for users. Because the residual amount of resourcesof each physical host is always changing, it can not be guaranteed that the task can bedistributed to the physical host which has the largest residual amount of resource. Aiming atthe current issue that the most physical hosts of the cloud data centers are overloaded whichresults in the imbalance of the cloud data center, this text will do in-depth study of theselection issue of deploying physical hosts.The method of task deployment has become a research hotspot of green cloud data center.Actually it is a research of the selection issue of deploying physical hosts. Because it is anoptimization problem, we need to find one or more of the evaluation criterion. At present,researches on deployment tasks mainly use energy conservation or load balancing asmeasurement. Whereas, what this text studies is the task deployment problem of the loadbalancing in data center. It aims to find a deployment method as follows: Firstly, it makes thefailure number of the strategy’s deployment tasks be relative minimum in all deploymentstrategy; Secondly, after accomplishing deployment tasks according to this strategy, the wholenetwork in the cloud data center has a relatively optimal load balancing degree; Finally, comparing with other deployment strategies, using this task deployment strategy not onlymakes the whole cloud data center have a relatively higher throughput, but also optimizesexternal service performance of the data center. The most important thing is that those threeaspects above-mentioned should be accomplished on the basis of ensuring deployment tasks’performance and efficiency. In view of this deployment strategy, this paper presents asensitive-load-balancing clustering deployment method research under cloud computingenvironment. First, according to the fitness function of physical host’s performance, it does aconstrain limit to all physical hosts in cloud data center. Thus, it can achieve a taskdeployment strategy which has global search capability. Then, we use modified clusteringalgorithm to achieve further optimization and get clustering result eventually. Furthermore,the whole method realizes load balancing of the cloud data center. The main work of this textis as follows:(1) Expound the research background and the significance of this article. Introducecurrent research status of load balancing of the cloud data center home and abroad.Emphatically introduce dynamic load balancing, static load balancing and research progressof live migration which influences load balancing. Analyze the advantages and disadvantagesof the related research in detail.(2) First, we introduce the basic concept and the architecture of cloud computing. Andwe introduce the concept of load balancing and some algorithms achieving load balancing.Finally, we introduce briefly clustering algorithm and CloudSim simulation platform whichare used in this text.(3) We give the model and system architecture of task deployment method and describeits logic execution process in detail. On the basis of this, we present asensitive-load-balancing clustering deployment method, LB-C (Load Balancing UsingClustering).LB-C is a heuristic task deployment clustering optimization algorithm based onimproving clustering algorithm. This paper gives the specific design and implementation ofLB-C. The task deployment strategy proposed can not only find the best physical host todeploy tasks, but can realize long-term load balancing of the cloud data center.(4) Through the contrast experiment, we effectively evaluate LB-C algorithm. Theexperimental result shows that compared with the existing research, LB-C significantlyreduces the failure number of deployment tasks, improves the cloud data center’s throughput,optimizes external performance service of data center, has better load balancing effect andmakes the cloud data center be more green and efficient.