As the kernel of the enterprise information system, DBMS is in charge of the enterprise's pivotal tasks and millions of users' intercurrent accessing. The low efficiency of the general commercial database management system in OLTP become the bottleneck of upgrading capability and efficiency of the enterprise information system. For the demand, many enterprises and research institutions put forward some solutions, for example, cache accelerate, memory accelerate, database cluster accelerate etc. For OLTP application, the database cluster is a effectual solution. The database accelerative engine which makes use of the database cluster technology is a common database cluster solution.The main idea of the accelerative solution is that the tasks which are charged with a machine are dispatched many machines, through data fragmentation and global transaction management, which have important meaning to enhance the accelerative effect.Data distribution is requested for balance. We must design a effective data fragmentation arithmetic to balance the load of nodes. The Data Distribution of the database accelerative engine includes two fragmentation technology: Hash horizontal fragmentation and vertical fragmentation, which can be applicable to OLTP nicely. As the kernel of the database accelerative engine, the global transaction management must insure the atomicity of the global transaction and the validity of concurrency control. Through setting the local agents at all nodes, the database accelerative engine insure the atomicity of the global transaction. At the validity of concurrency control, we improve on two level serializability standard, and put forward a new transaction commitment arithmetic.TPC-C test prove that the database accelerative engine has high price performance ratio and improve the power of parallel processing, a certain extent advance the executive performance of OLTP. |