** Traditional database performance issues **
Traditional databases refer to traditional relational databases such as SQL SERVER, ORACLE, MYSQL and DB2. Its performance follows the following principles: Traditional databases strictly adhere to ACID properties, commonly understood as the pursuit of ‘strict and accurate data’, so its horizontal scalability (that is, multi-machine distribution to improve the concurrency and data volume of the whole system) is relatively poor, or it is’ maximum ‘, that is, there is a ‘ceiling’. Horizontal scaling (distributed systems with tens, hundreds, or even thousands of servers) is not as easy as the current popular NOSQL-type databases. This is determined by the nature of traditional databases, the so-called ‘gain, lose’. Second: due to the constraint of the first feature above, when the amount of concurrency or data is large, the only way is to separate the database according to the type of business, so that different businesses run in an independent database system on an independent physical database server. Third: improve the performance of database system by upgrading hardware and software optimization. Fourth: use DBTWIN cluster load balancing to improve the overall performance of the database system. Because DBTWIN load balancing is for query operations, its based on real-time data synchronization consistent, so the system hardware level impact on the performance of the whole cluster is critical, especially the network hardware class, MB, gigabit, Wan Zhao network’s influence on data synchronization difference is very big, even great Yu Wanzhao Ethernet high-speed network speed, This configuration is necessary for a system that pursues high performance and high concurrency. (The faster the network speed is, the closer the performance of DML operation is to that of a single node, thus the performance of the entire DBTWIN cluster is significantly improved.) Fifth: use SQL own replication, mirroring, ALWAYSON and other asynchronous data transmission way to achieve the separation of OLTP and OLAP. Due to the data time difference between the primary and secondary nodes in this way, manual processing needs to be combined with the code, and the management and maintenance is complex, and there is a hidden danger of data logic error, so it is not universal.