Managing the Physical Log

The PHYSFILE parameter specifies the size of the physical log. Because a checkpoint occurs whenever the physical log becomes 75 percent full, this parameter affects checkpoint intervals.

To calculate the number of kilobytes required by the physical log if your workload requires intensive updates and updates do not usually occur on the same data and index pages, use the following formula:

PHYSFILE = (connections * 4 * max_txs_per_user * pagesize) / 1024

When you estimate or adjust the size of the physical log, consider the following factors:

• For fast recovery after emergency shutdown of the database server, specify a physical log size that is large enough to contain about twenty-five percent more pages than the total number of physically logged pages for two checkpoints. To calculate this number, display the checkpoint history statistics. Execute the onstat -g ckp command.
• If applications require fewer updates or if updates tend to cluster within the same data pages, you might reduce the size of the physical log.
• If you expect increased activity or increase the checkpoint interval, as explained in Specifying the Interval Between Checkpoints, increase the size of the physical log.
• To use physical-log fullness to trigger checkpoints, decrease the size of the physical log.