Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: trace analyzer / tkprof

Re: trace analyzer / tkprof

From: danisment <danisment_at_yahoo.com>
Date: 19 Jun 2003 03:26:03 -0700
Message-ID: <2c78cfac.0306190226.68c54521@posting.google.com>


Hi,

I've not documented, but according to my tests, Oracle doesn't update some statistics for X$ tables. Which SQL are you using ?

btw,

Be careful while interpreting raw trace files. Because, some statistics(not all) include statistics of their recursive statements' statistics. So, it's duplicated in raw trace files like in Oracle dictionary. This is really a problem if a long running statement calls a nothor long running statement. For example:

begin
 select ....
end;

if select statements take 1 hour to complete, you will see about total of 2 hours for begin..end block, and insert statement.

I don't know other products, but Oracle's tkprof, and ubTools' itrprof substruct recursive statistics from parent statements and returns excat values f&#305;r statements.

best regards...

Danisment Gazi Unal
http://www.ubTools.com
Web Based Oracle Products and Services

"Oradba Linux" <oradba_linux_at_attbi.com> wrote in message news:<4QQHa.14592$hz1.21438_at_sccrnsc01>...
> I was reading about trace analyzer , a tool used to mine raw trace files and
> provides more information than tkprof does . I am
> thinking of trying it out to some trace files and learning more about it .
> Comments about trace analyzer are welcome .
>
> Meanwhile i was looking at the output from trace analyzer , it is just an
> example from a support note . Can i ignore the count
> zero time in the below report . I think it is a count of waits which had
> ela=0 in the raw trace file .
> Does (Total Time waited / ( Times Waited - Count Zero Time) ) tell us the
> time taken for an I/O operation to complete ?
>
> Event Times
> Count Max. Total Blocks
> waited on Waited
> Zero Time Wait Waited Accessed
> ----------------------------------------------------------------- ---------
> --------- --------- --------- ---------
> db file sequential read (single block index scan)................ 521
> 284 0.05 2.84 521
>
> Also in one of the raw trace files
> FETCH
> #<CURSOR>:c=0,e=0,p=0,cr=0,cu=0,mis=0,r=1000,dep=0,og=2,tim=xxxxxxxxxxxxxxx
> I see a fetch operation with p=0,cr=0,cu=0 but rows processed = 1000 . How
> do i interpret such a situation ?
>
> Thanks
Received on Thu Jun 19 2003 - 05:26:03 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US