Re: STATSPACK Report Accuracy???

Statspack is simply reading v$filestat,
and v$filestat (and v$tempstat) can
acquire meaningless timing results.

Generally, anything which results in async, or pseudo async activity can lead to
ridiculous times in in v$filestat. (e.g. multiple db writers, Oracle I/O slaves,
O/S async I/O). On occasion I have
seen timing results where the I/O time
seemed to be proportional to the file id.

If you want to cross check, look at the
values (particularly the max wait time)
on v$system_event/v$session_event for
event "db file sequential read". For some odd reason, these seem to be realistic even when the times on v$filestat are insane.

If you want to take regular snapshots, you can use dbms_system.kcfrms to reset the
max times.

-- 
Regards

Jonathan Lewis
http://www.jlcomp.demon.co.uk

The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/faq/ind_faq.html

Optimising Oracle Seminar - latest dates
http://www.jlcomp.demon.co.uk/seminar.html



"Charles Davis" <cdavis10717_at_comcast.net> wrote in message
news:4cadnR3Xs8IqSgrdRVn-ug_at_comcast.com...


> When running STATSPACK reports I see some Datafile I/O Avg Read Times of

> 10's of thousands of ms.  Such as 85,000 Avg Read Times in millisenconds.

> Wow!!

>

> Is STATSPACK accurate?  can it be i have Oracle waiting 85 seconds on

> average to read a block from a file?

>

> With each 24 hour snap I see similar and higher numbers, but always on

> different files.

>

> Am beginning to think a problem with the storage array...

>

> All disk files are on EMC disks using PowerPath an SRDF.

>

> Appreciate anyone's opinion or feedback.

>

> thanks.

>

> Charles

>

>