What you really need is to simulate your production traffic in your SAP stress testing environment. This is a very big task, involves a lot of application support and script writers. Here in-house we use LoadRunner. Why do this? Because if you look at ST04, the record there will only be accurate if you can guarantee the business transaction going through the system is production-alike.
For example, if you have a specific project dealing with Finance that's going live, if you do not have a stress environment that simulates production, what you'd end up w/ is a tuning based on strictly Finance, w/ out consideration that maybe Finance is only a small % of your real life traffic.
Once you have a proper stress testing suite setup, give it a run and then check ST04 and DB02 for your table statistics. Use perfmon to monitor paging, CPU, storage I/O, etc.
Lastly, to guage how accurate your stress testing environment is, you can compare the ST04 and DB02 to production. If the top n tables, top growing tables, etc. are about the same, then it's pretty accurate.
As for missing indexes: This can be handled a few different ways. Primary way of handling this is reactive-one-at-a-time mode. Meaning, running ST05 for each slow running program and tuning it. Second way is ST04 > detailed analysis > SAP stats on SPs. Look at the slowest running programs, tune them individually.
For SQL 2005, there are also index / performance statistics. But remember, these tables are cleared whenever SQL engine is cycled. So you may want to run a collector SQL job that collects the stats into a permanant table, and maybe run this collector once every 15 minutes.
Edited by: Kevin Lin on Jul 1, 2008 12:00 AM
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.