Archives of the TeradataForum

Message Posted: Tue, 25 Jul 2006 @ 21:33:31 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

<-- Anonymously Posted: Tuesday, July 25, 2006 17:17 -->

Original question:

Here is the deal...

Tereadata is design to use all available resource so output from sar indicate 100% usage may not necessary be a indication of performance impact to the system overall.

Sar output may not necessary be the best information to judge how your system is performing. Please follow the few chapters regarding to trouble shooting performance problem in Performance Management Reference manual.

Usually you should see if you have database impact by using schmon -m to see the wait queue time frame is acceptable for each of the priority (workload) groups. You should see the result from exec dbc.resnode(date-1, date, 0,235959); if you keep your resource usage data in database DBC. You will at least see the data for every 10 minutes instead of every hour. If you have heavy duty job scheduled to run every hour on the hour then you will be mis-lead.

You should see how your PSF was setup and whether user from the groups with SLA was impacted. You should gather information in DBQL and AMPUSAGE to help identify the resource hogs.

You should have canary setup to determine whether you have system level impact or application level impact by the processes that consumed 100% of the cpu resource.

There are also IO resource you are not looking at through sar which schmon -m will provide.

Anyway, follow the suggestion in performance management guide and if you need help from the vendor.

	<Prev	Next>		<<First	<Prev	Next>	Last>>