Page Grouped and Reverse-ordered Redo logging to accelerate recovery for database system
Publication Date: 2015-Jun-09
The IP.com Prior Art Database
Unlike what most database system do, this article introduce a new method to organize redo logs records on physical log files to speed up table recovery in the future.
Page 01 of 3
Page Grouped and Reverse
Page Grouped and Reverse-
--ordered Redo logging to accelerate recovery for database system
ordered Redo logging to accelerate recovery for database system
It is a vital function to recover one table to current or point-in-time in various database systems, no matter it is relational or non-relational. Normally it implements recovery process by applying redo logs. However there is problem that we may read a great number of unnecessary log records for recovery since redo and undo logs are mixed together in traditional logging mechanism. An prior art introduced an method to separate redo and undo logs that can deal with this problem. However some other issues still remain: 1. Redo logs for different tables are still mixed so it reads all redo logs despite we only intend to recover one particular table
2. Database users may issue multiple recovery commands in unpredictable time, then concurrent read for redo logs could lead to contention, which degrade the recover performance.
3. Recovery is usually based on restore of image copy which is in ascendant page order. So basically the "big pages" should be in local buffer after restore finishes. If we first apply redo log for "small pages", it needs extra I/O read to get "small pages" into buffer again.
The core idea o this invention includes:
1. Group redo log by table space and table pages so that concurrent recoveries on different tables can deal with their own log segments.
2. Reverse page order for redo log to save I/O read in applying log ph...