Research On Key Technologies Of High Performance Database Cluster In Cloud Environment

With the development of cloud computing technology,it has been widely used in all walks of life.As a key component of the cloud platform,the database system has an important influence on the performance of the cloud platform.Therefore,exploring a higher performance database system is a key problem concerned in the industry.Compared with the stand-alone database,the database cluster,which has high availability,high performance,data backup and other advantages,has become the preferred platform for building the cloud platform.Facing the complex cloud environment,the traditional database cluster cannot solve the problem of sudden increase of traffic,and the algorithm of load balancing is also lackluster.In addition,some database clusters also have strict requirements on hardware and high licensing fees,which is contrary to the idea of cloud computing.In this context,this thesis studies the key technologies of high performance database cluster in cloud environment,and implements a database cluster solution based on MySQL.The solution effectively leverages the cheap machines in the cloud platform,reducing hardware and software costs and improving overall cluster performance and throughput.First of all,the proxy module is added in front of the cluster.The proxy is transparent to the users so that they do not perceive the existence of the proxy,and users access the agent just as if they access the database as normal.The proxy implements the single system image of the database cluster and SQL can be routed and forwarded by proxy.Therefore,you can use the proxy module to achieve the separation of read and write.Then,this thesis analyzes the shortcomings of the existing load balancing algorithms and proposes a new one.The algorithm determines the weight of each resource on server performance by AHP,and quantifies the load capacity of the server to establish the foundation for load scheduling decision,and improves the allocation decision with probability method.At the same time,the algorithm also considers the overload problem of single resource on the server,and resolves it by putting the threshold limit on it.On the basis of MySQL Replication,this thesis designs a data replication scheme which processes the data replication hierarchically by the means of cascade.A dynamic detection algorithm based on linked list is proposed for cascade fault caused by hierarchical replication.In order to ensure the consistency of data access,the solution of this thesis is to check current replication state of the Master node.Then node waits for data copied from the parent node.It does not execute the user's request until the node's state is similar to the Master node.Based on MySQL,this thesis implements a database cluster system which the previous design ideas are applied in.Finally,this thesis designs a series of experiments to test the effect of load balancing algorithm and the read and write performance of the cluster,which proves that the performance of the cluster is improved.