Database Management System is one of the most complex software systems.Inorder to support the ACID (Atomicity, Consistency, Isolation, Durability) ofconcurrency transactions, high concurrency and high performance, DBMS mustmaintain the serializability of transactions by using a kind of concurrency control.Thedispatcher is in charge of it.Generally speaking, there are three dispatchers and they arelock based, timeatarnp based and validation based respectively. As a relation DBMSsupporting transaction completely, IrmoDB data engine of MySQL uses the dispatcherbased on lock. InnoDB lock system has the traditional two-phase locking policy systemand can make row level read-write lock. This paper makes an in-depth study on thelock-based concurrency control mechanism of InnoDB data engine of MySQL,including lock type, lock model and lock stucture, which lays good foundation for therealization of MySQL self-optimization, especially for the realization of concurrencycontrol self-optimization.Abstract Selfmanagement is a currently critical problem of database managementsystem. The key solution to selfmanagement is to make database system manage itselfand use its resources effectively. This paper realizes two self-optimization DBMSmodels in the perspectives of concurrency control and memory resource respectively. Theformer will make your database system more intelligent powerful and costless bypredicting and controlling the dam conflict ratio. The later can adjust system parametersautomatically to achieve good performance. In addition, both the two models haverealized obvious optimizing effects by means of experiment testing. |