Write Ahead Logging With Efficient Restart
Original Publication Date: 1984-May-01
Included in the Prior Art Database: 2005-Feb-06
This relates to a method for write ahead logging with efficient restart in a CPU system supporting any data base. The method comprises the steps of (a) modifying pages and logging of a description of said page modifications; and at the time of failure (b) identifying which transactions were incomplete (needing undo), identifying unforced pages modified in a buffer pool (needing redo) by scanning forward along the system log starting at the last checkpoint until a null or restart log record is encountered and building transactions and dirty page tables; (c) executing "undo processing"; and (d) executing "redo processing". This method may be used for both record level as well as page level locking.