The Research Of Load-balancing Algorithm Based On OpenStack

Cloud computing is the integration of traditional computer technology and network technology,such as distributed computing,parallel computing,network storage,load balancing,etc.The main idea of cloud computing is to integrate all kinds of computing resources on the Internet,providing on-demand services to customers in the form of computing resources,so it is a new generation of an important technology of the information technology.With the rapid development and maturity of cloud computing technology,the scale of data center also increases dramatically.The quality of cloud service is directly determined by adjusting the load balance of cluster servers,reasonably scheduling tasks and allocating resources.OpenStack is saves as one of cloud computing open source projects,which provides a complete infrastructure as a service(IaaS)solution through various complementary services.Therefore,the research on cluster load balancing algorithm based on OpenStack is great significance for the overall performance and resource utilization maximization of cloud computing system.We research cloud computing and load balancing technology,and find that the traditional load balancing algorithm cannot be well applied in the cloud computing environment,and there are problems such as long system response time,load imbalance,and the server cluster is prone to load skew when the concurrency is large.In addition,in the reading strategy of the Swift component,when the file is read,the agent node is responsible for randomly assigning tasks to one of the storage nodes.In this case,imbalance of disk utilization is likely to occur,leading to load imbalance.To solve the above problems,two algorithms are proposed as.follows in this paper.(1)An improved grouping weighted minimum connection grouping scheduling algorithm is proposed.The algorithm is based on the hardware parameters of various resources according to RS.Accurately calculate its weight;then consider the difference of the tasks to be processed and the outstanding advantages of the RS in dealing with different tasks,and RS is grouped to ensure that the number of RSs in each group is basically the same;then RS periodically collects its own memory,CPU,I The utilization of resources such as I/O issent to the LB for dynamic adjustment of the RS weight;the round-robin scheduling algorithm(RR)is used between the groups,and the WLC algorithm is used between the groups.After the grouping,the number of servers to be managed by the LB is reduced,and the groups are parallelized.The calculation greatly improves the efficiency.Experiments show that the proposed algorithm is more effective than RR and WLC.(2)A method of sorting storage nodes is proposed to design a reading strategy,so that the task of file reading can be more balanced allocated to each storage node to achieve load balancing.The experimental results verify that the improved strategy of this paper is more advantageous than the Swift inherent strategy in balancing the CPU utilization,I/O utilization and network outflow rate of each storage node.