- Workload
- Class 1-2 relationship
- Class wait time
– Components
– Average 1/O times
– Total % of class 2 - Deadlocks
- TimeOuts
- Buffer Pool hit ratio(s)
– Application hit %
– System % (GP PagesRd)/GP - HP Retrieval Ratio
– PgsReadback/PGSWritten - Rid Pool Failures
– How many
– Why - Online Transaction – Seq. scan
You must always consider the amount of work your application is performing – look at the number of getpages, fetches, I/Os, locks, etc, etc.
Look for the problem areas. What can be improved?
Calculate the System BP hit ratios, calculate the avg. I/O wait time by dividing the total synch wait time by the number of I/Os.
Look at any failures of resources, and look for things that don’t seem normal to you.
Much of this can be programmed into your online monitor exception reports, or other reporting facilities. You really want a reasonable level of exception reporting so you don’t have to look at tons of detail reports.