Research Of Data Locking And Consistency In Distributed Database

With the development of network technology, the application of distributed database is becoming increasingly important. Based on Hunan science and technology planning project, this thesis has studied locking algorithm and the consistency control algorithm in distributed database and has explored the distribution and the redundancy of data, the distribution of transaction execution and some other unexpected problems that are caused by software and hardware faults. Faced with these problems, the present study has made a lot of research and put forward some useful solutions on the basis of the existing consistent algorithm in distributed database. The content of this study is shown as follows:This thesis has studied the simple distributed locking algorithm and the full distributed locking algorithm. As these algorithms have many demerits such as high expense in communication, long locking time and complex locking management, this thesis brings forth a locking management algorithm based on global directory in distributed database in the basis of locking type in central database, locking granularity, global directory and three-stage locking protocol in distributed database. In this algorithm, global directory sever is responsible for maintaining all locking information and global locking manager is in charge of managing and maintaining locking node information of global directory.This paper has made a study of the data copy technology of distributed database and explored the demerits of maintaining data consistency which exist in message queue algorithm, transaction control algorithm and copy control algorithm. Therefore, the study has come up with data consistency algorithm based on global directory in distributed database. This algorithm can realize the asynchronous updating, which can save network bandwidth and improve the transaction process.Using Oracle 10g, my Eclipse and some other tools, the artificial system of data locking and data consistency in distributed database has been made. Then the test of transaction press has been made by SysBench. And the whole test process has been supervised by the database management tool Toad. The research result indicates that the two algorithms mentioned above are correct and feasible.