02-13-2007 11:15 PM
Hi
Does somebody knows where can i find help for understand what is or how to 'read' the Estimated Cost Clause for the Selects in the trace ST05 analysis.
Regards
Frank
02-14-2007 2:46 PM
I use the estimated costs only for comparison purposes - for example two different SELECTS trying to get the same data.
Rob
02-14-2007 2:46 PM
I use the estimated costs only for comparison purposes - for example two different SELECTS trying to get the same data.
Rob
02-14-2007 6:49 PM
Hi Rob
That's just what i want to do, compare one original select with other modified select wich get de same data (this new select replace the first) for compare wich one is better for performance.
Is there some help for can do this comparison using the ST05 Estimated Cost ?
Regards
Frank
02-14-2007 6:58 PM
So if you run both SELECTs, you should get two different estimated costs. The plus side of this is that it is unaffected by any buffering. The down side is that bot selects may be translated into a different number of actual database SELECTs. This makes the comparison difficult.
I prefer a more "real world" approach and instead of ST05, I use either GET TIME or GET RUNTIME multiple times against both SELECTs.
Rob
02-14-2007 7:16 PM
Thanks Rob
That's good point, i was trying to use the ST05 for compare two selects, but just like you said, the ST05 time is affected by work-load server and the Estimated Cost is desglozed in way not easy to 'read'.
So, talking about use GETTIME or GETRUNTIME, what's your suggestion; use these for each select: one first time begining the session, then anhoter time for take in account the buffering, then anhoter time (3 times for each Select) ?
And just for confirm, GETTIME and GETRUNTIME are affected too by work-load server ?
Thanks a lot.
Frank
02-14-2007 7:56 PM
This may help:
/people/rob.burbank/blog/2006/09/13/using-an-index-when-you-dont-have-all-of-the-fields
Rob