The Research Of Load-balancing Algorithm Based On OpenStack

Cloud computing is not only integration of both traditional computer technology and network technology,but also vital strategic technology and means that will lead innovation of future information industry.As cloud computing is becoming more and more popular,it is more urgent than ever to improve the quality of cloud service,the utilization of IDC resource and reduce the energy consumption,customer cost at the same time.It's of great importance to do a series of further research into this subject.In the cloud computing clusters,the problem which must be addressed is the load balancing of cluster nodes.In order to deal with the new issue caused by the application of the traditional load balancing techniques to the cloud computing environment,based on the most popular open source cloud to computing platform OpenStack,a series of research was launched.First,this thesis introduces the field background of both domestic and foreign researches of cloud computing technology and load balance,and summarizes the related theories and concepts,analyzed current the basic characteristics of clouding,the key technology and classification,comparing with several mainstream open cloud computing platform.The focus of this study is using OpenStack to build a cloud computing architecture which is based on Infrastructure as a Service(IaaS),OpenStack components and logical architecture to do in-depth research,and create the virtual machine.Second,by studying the load-balancing algorithm,a new algorithm of cluster groups is proposed.It calculates the weight of each node at first and makes the servers into different groups in accordance with their performance which should be equal(or approximate equal).The quantity of nodes in each group should also be equal(or approximate equal).Next,LB regularly collects the data of CPU,Memory and so on,and dynamically alters the weight of nodes,and selects the best node with Round Robin(RR)in this group.At the same time,calculate the comprehensive load of each node and group load.Finally,using improved Weighted Least Connections(WLC)algorithm selects the best node of the cluster and carries out the rational distribution of task requests via the Group Load Balancer(GLB).As a result,this algorithm is better than RR and WLC.Third,according to the study of Cinder and deep analysis the metadata of Ceph management strategy based on directory migration dynamic load balancing strategy,and then considering the heterogeneous of RS,CUP and Memory,an improved scheme is proposed,which based on the request time of each node.As a result,this algorithm can make the metadata management strategy more effective.