Two types of user-specified durations may be associated with exceptions:
- A threshold. The specified action occurs as soon as the threshold is exceeded. Most exceptions are in this category. These include the following exceptions:
- Maximum spool rows
- Spool usage in bytes
- I/O count
- Spool size
- Blocked time
- Elapsed time (with the options to exclude blocked time and exclude throttle delay time)
- Number of AMPs
- CPU time (summed over all nodes)
- Tactical CPU usage threshold per node
- Tactical I/O physical bytes per node
- I/O physical bytes
- A threshold and a qualification period. The specified action occurs only after the exception has persisted beyond a certain value for a specified time. The following exceptions are in this category:
- CPU Disk Ratio
- Skew (I/O or CPU skew difference or skew percentage)
Exceptions can have one or multiple criteria. When an exception has multiple criteria, all criteria must be exceeded to trigger the action.
Select values for exception criteria based on typical values for requests in that workload. Identify boundaries that requests in that workload should not cross. For example, if the typical request in the Finance workload takes about 250 CPU seconds, and almost all requests in the workload range from 100 to 1,000 CPU seconds, any request that exceeds 1,000 CPU seconds is out of place in the Finance workload and is running at the incorrect priority. In this case, you might create an exception to take automated action on any request that exceeds 1,000 CPU seconds.