Database Cluster Load Balancing Technology

With the rapid development of e-commerce and information technology and wide application of Internet, database is widely applied in different industries and fields. A flood of information and a large of user requests place a server challenge on database management system. The database cluster provides a solution to solve such problem encountered by traditional database management system. However, in order to make database timely and efficiently provide users high-quality services, under the premise of low-input, the question that how to allocate tasks to make sure the distribution of tasks and loads is reasonable according to performance of nodes and the exiting load and the question that how to improve the reliability of the cluster and scalability of the cluster must be solved. Load-balancing technology can help us to solve this problem. Through the load-balancing technology, we can allocate new requests according to the load information of back-end node, at the same time, we can also adjust the operating requests of user to improve the system response and meet customer's request.This paper firstly introduces the research background and significance of the database cluster and basic theory and key technologies of the database cluster system. And then designs a middleware structure that can help to realize load-balancing algorithm, analyses the function of each module of the middleware structure and introduces the key technologies to achieve each module.On the basis of detailed analyzing the factors that affect load-balancing, a dynamic load balancing algorithm is put forward. The algorithm can effectively achieve the collection of load information, evaluation and placement; also can effectively balance the load of server node to improve the efficiency of the system.The specific work done is as follows:1. On the basis of studying database cluster and middleware, we design a architecture that can help us to run load-balancing algorithm, and analyze the composition and function of the architecture. The structure divides user's requests into smaller requests that are assigned to the nodes by two transaction manager and enhance the collaboration during nodes through the reasonable division to the user's request.2. By analyzing the application request of database and the characteristics of data distribution and heterogeneity of nodes, we design a data distribution that base on replication and one-dimensional view data partition. When the data distribution is converted, it can avoid transferring large amounts of data. At the same time, we design a dynamic load-balancing algorithm on this basis, which can improve the write operations of database more efficient under the premise of maintaining the high efficiency of read operation. 3. Finally, we put this load-balancing algorithm to experiment. The results prove that this algorithm realizes the expected goals.