Estimating after-imaging disk space requirements

Before creating after-image files, it is important to accurately estimate the amount of disk space required to hold all the AIdata for your database. Even when you use variable extents, if the disk runs out of space and there is no empty AI extent available, you must perform emergency maintenance or the database is forced to shut down. See Recoveringa Database for details. To prevent the database engine from shutting down when it exhausts AI disk space, start your database with the after-image stall (-aistall) startup parameter. When after-image stall is enabled, the database will suspend all activity, and send an error message to the log file indicating that the AI extent is full. For more information on after-image stall, see After-imageStall (-aistall).

To determine the amount of AI file space required, use the PROMON utility's BI Log Activity Display option to monitor the number of BI writes that occur during a period of time. The information that is written in the BI file closely matches what is written in the AI file. Typically 0.05 percent fewer bytes are written to an AI file than are written to the BI file. Therefore, the BI Bytes written statistic provides a close estimate of the space required for an AI file for the same activity period.

Measure the BI Bytes written statistic several times toward the end of the activity period that you are sampling. If your workload varies from day to day or peaks during end-of-month processing, count the BI writes during a peak period to ensure that you allocate enough space to handle peak days. If the workload grows, make sure you increase the amount of space available for AI files. Always have extra disk space available for unanticipated growth of the AI file.

Another way to calculate your AI file size is to use Before-Image Cluster Size from the PROMON utility BI Log Activity Display and the Checkpoints from the Activity Display. If you want your AI file to hold all the changes made over a two hour period, determine how many checkpoints occurred during a two hour time period of peak activity. The number of checkpoints multiplied by the BI cluster size equals the minimum size your should make your AI extents to capture the changes. For example, if you have four checkpoints in two hours, and each cluster is 0.5MB, then you need your AI files to be at least 2.0MB to record two hours of activity.

Once you have sized your after-image areas, you need to manage them so that there is always space available to write in an extent. See Managingafter-imaging files for details on managing your after-image areas.