Design And Implementation Of Relational-Database System Transaction Recovery Strategy

For guaranteeing the security and consistency of data in the failure of DBMS, DBMS preserve some redundancy data, include backup and log information. As a large-scale object-relational database, PostgreSQL has adopted both of the information. PostgreSQL record the history of update of database in security mode, so that the system can recover from a crash. This dissertation analyse the implementation of the transaction manager, log manager and recovery manager, we bring forward a schema to improve the capability of log manager and recovery manager, and implemented it on PostgreSQL. According to the result of test, we briefly introduce the research directions and point out the future methods to improve the capability of PostgreSQL transaction processing. The contributions of the dissertation are followings: 1) Minish the content of PostgreSQL's log greatly, improve the capability of transaction processing, by way of adopting doublewrite buffe technology. 2) Append the flush link in the data buffer pool and startup daemon writer process, reduce I/O cost during checkpoint, so that lessen the influence of the performance of DBMS. 3) By way of analyzing the result of test, we summarize some factors of enhancing the capability of transaction processing, and bring forward some new schems. According to the dissertation, we amend the performance of PostgreSQL transaction processing and develop an improved prototype based on the PostgreSQL.