Can I accept the AVGIOTIM colum as accurate then? BTW does you know if
the optimizer uses this number in choosing an execution plan?
--
Chuck
jhunter_at_iDevelopment.info (Jeffrey Hunter) wrote in
news:9dfc69b4.0304221401.506a89a_at_posting.google.com:
> Hello Chuck:
>
> I took a very quick look to see if there were any known issues with
> v$filestat and here is what I found. Not sure if this helps or not,
> but I come up with similar discrepancies between calculated and
> avgiotim. Tested on: 8.1.7 on Linux and 9.2.0 on Solaris.
>
> Known Issues:
> The times in the WRITETIM column are often incorrect. If the
> average
> write time (WRITETIM/PHYWRTS) looks nonsense then it probably is.
> Incorrect values typically show up as about 10 times more than one
> would expect.
>
> Regards,
> -- jeff
>
> Chuck <chuckh_at_softhome.net> wrote in message
> news:<Xns936598D24D465chuckhsofthomenet_at_130.133.1.4>...
>> Can someone explain why the avgiotim column from v$filestat doesn't
>> match the time computed by dividing total i/o's by total i/o time?
>> Timed statistics is set to true. The platform is AIX 4.3, Oracle
>> 8.1.7.
>>
>> col avgiotim heading "AVG|IO|TIME"
>> col tot_io_time format 99999999999 heading "TOT|IO|TIME"
>> col computed_avg_io_time heading 'COMPUTED|AVG IO|TIME'
>>
>> SELECT FILE#
>> , avgiotim
>> , phyrds + phywrts tot_physio
>> , readtim + writetim tot_io_time
>> , (readtim + writetim) / (phyrds + phywrts) computed_avg_io_time
>> FROM v$filestat
>> WHERE file# in (4,14)
>> ORDER BY 3 DESC
>> /
>>
>>
>> AVG TOT COMPUTED
>> IO IO AVG IO
>> FILE# TIME TOT_PHYSIO TIME TIME
>> ------ ---------- ---------- ------------ ----------
>> 14 0 117405621 10124912412 86.2387365
>> 4 4 320 422 1.31875
>
Received on Thu Apr 24 2003 - 09:39:05 CDT
