Design And Implementation Of Amoeba-based Database Load Balancing Middleware

Based on Amoeba which is an open-source middleware developed a database load balancing middleware which applies to high concurrency, large amounts of data. Transparent to the user, while independent to the business systems and the database systems. After deals the user requests with filter resolution, connection limits and other treatment, use the load balancing algorithms and literacy acquisition separation to process the request on the target database server. And finally the results of the integration of user requests back to the user.When selecting the load balancing algorithm, in order to apply better to caching mechanism application database cluster system, consistent hashing algorithm is introduced in this middleware. Depending on the number of physical nodes configure the appropriate number of virtual nodes, and these nodes composition to a hash calculated over the formation of a hash ring. According to the characteristics of the first hash ring connected to find the first available node in the clockwise direction. And this can significantly reduce the bad impact to the cluster when added or removed nodes to cache system. In order to make the results of hashing algorithm more uniform, consistent hashing algorithm applied in the Murmur Hash algorithms which has excellent performance in stable, capability and so on aspects. Read/Write Splitting technology has the ability to distributed, written or read data requests, the two distinct data requests are distributed to the corresponding database. Meanwhile the technology can be configured to ratio between the write database and the read database according to the actual situation of business scenarios, to further improve the database cluster data processing capabilities. Apply My SQL replication mechanism to achieve master-slave synchronization, solving data inconsistencies which is caused by Read/Write Splitting technology.The analysis from the results, this database load balancing middleware achieved load balancing, Read/Write Splitting, connection limits and so on functions. And the use of My SQL replication mechanism to achieve a master-slave synchronization to meet the business needs. Meanwhile, the middleware solved database performance data processing bottlenecks.