Re: Query Timings

I sometimes look at that "raw" *.trc file. It is perfectly readable ASCII. The tkprof program aggregates (sums) all the different instances of the same SQL statement. This is OK if you want the sums of the various measures, disk reads, elapsed time, etc. but not if you want to find the maxima.

I am hoping that Raju (or someone else) can clarify the meaning of the various letters used to identify the statistics. A typical line looks like sort of like
the following (but I removed the = signs and commas to increase readibility). FETCH #86:c 18 e 60 p 65 cr 1271 cu 0 mis 0 r 200 dep 0 og 3 tim 28655380

There are separate lines for PARSE, EXEXCUTE, and FETCH. #86 is the cursor number.

I believe that c is the CPU time and e is the elapsed time, both in hundreths of seconds. I think that p is physical disk reads (corresponds to "disk" in tkprof) and cr and cu correspond to query and current, and that r is rows.

I believe that tim is the time, i.e. seconds since the Unix epoch started 1/1/70. True?

Can someone please clarify the meaning of mis, dep, and og.

In article <4cjrbn$6m0_at_gw.PacBell.COM>, Raju Patel <rpatel_at_hooked.net> wrote:
>You need to set SQL_TRACE=true for that session or globally in the
>init.ora. For the session, use the command ALTER SESSION SET
>SQL_TRACE=true. You must also have TIMED_STATISTICS=true in the
>init.ora. Then, when you run a query, it will generate a trace file in
>the user_dump_dest directory. This is a raw trace file. You will then
>need to use tkprof to format the trace file. In 7.1, look at the
>...

-- 
Steven Tolkin	       steve.tolkin_at_fmr.com    (617) 563-0516   fax 476-9732
Fidelity Investments   82 Devonshire St. A5B   Boston MA 02109
Any opinions are my own.   I'm not really an actor -- I just play one on TV.