Syntax options:
FAST COMMIT IS ENABLED | FAST COMMIT IS DISABLED
By default, Oracle Rdb writes updated database pages to the
disk each time a transaction executes the COMMIT statement. If
a transaction fails before committing, Oracle Rdb only needs to
roll back (undo) the current failed transaction; it never has to
redo previous successful transactions.
You can change the commit processing method by enabling journal
fast commit processing. With journal fast commit enabled, Oracle
Rdb keeps updated pages in the buffer pool (in memory) and does
not write the pages to the disk when a transaction commits. The
updated pages remain in the buffer pool until the process meets a
condition specified by the database administrator or applications
programmer. At the moment the condition is met (the checkpoint),
all the pages the process updated for multiple transactions are
written to the disk.
You can set a checkpoint for your process when:
o A fixed number of transactions are committed or aborted. You
set this by specifying CHECKPOINT EVERY n TRANSACTIONS.
o A specified time interval elapsed. You set this by specifying
the CHECKPOINT TIMED EVERY n SECONDS clause.
o The after-image journal (.aij) file increased by a specified
number of blocks. You set this by specifying the CHECKPOINT
INTERVAL IS n BLOCKS clause.
If a transaction fails, Oracle Rdb must undo the current, failed
transaction and redo all the committed transactions since the
last checkpoint. Redoing updates involves reading the .aij file
and reapplying the changes to the relevant data pages.
Fast commit processing applies only to data updates: erase,
modify, and store operations. Transactions that include
data definition statements, such as create logical area or
create index operations, force a checkpoint at the end of the
transaction. If you do not specify values with the FAST COMMIT
clause, the default values are applied.
NOTE
To enable FAST COMMIT, you must first enable after-image
journaling.