RE: cpu time and query column in tkprof output

Hi all,  

>Remember that your test results may differ from what you would see in =
>production due to differences in sessions, data volume, data distribution,
statistics, =
>index/table/view definitions, synonyms and more. Also, as someone
mentioned earlier, =
>latch statistics are counted globally so that when you are testing, you
need to
>realize that the latch stats you capture include all latching occurring
during your

I think we need to understand/remember one thing: That a 'latch free' wait will be recorded *only* when a session goes to sleep waiting for a latch. When the session spins, producing 'busy-wait' consumption and not really useful work, this is recorded under 'DB Time' (9.2+) in v$sesstat, and not under any latching stats. Thus, a test where some latching is involved (on account of its multiuser character) but is not severe enough to induce latch *waits* can give a tester the false sense of comfort. When using the wait interface exclusively, we tend to sometimes ignore this non-wait component and thus get into hot water when it moves to production and does not scale.....

Hth,
John Kanagaraj <><
DB Soft Inc
Phone: 408-970-7002 (W)

http://tahiti.oracle.com - Manuals for DBAs (English only) http://www.bibleserver.com - Manual for Life (in English, Deutsch, French, Italian, Spanish, Portugese, Turkish,...)

• The opinions and facts contained in this message are entirely mine and do not reflect those of my employer or customers ** -- http://www.freelists.org/webpage/oracle-l