Re: Significant difference in response times for same query running on Windows client vs database server

On Sunday, February 2, 2014 12:08:10 PM UTC-5, Mladen Gogala wrote:
> On Sat, 01 Feb 2014 09:00:42 -0800, Mark D Powell wrote:
>
>
> > mjm, you should capture the session that is running from the client and
>
> > compare the plan in v$sql_plan to what you posted. Why is the SQL using
>
> > a First_Rows hint?
>
> > The fact the client has a slower response time can be a sign the problem
>
> > is on the client machine: how many applications are running on the
>
> > client, how busy are the machine cpu's, how much free memory is there?,
>
> > and what is the speed of the machine cpu's? Checking to see if other
>
> > end users have better, worse, or equal performance may be worthwhile.
> >
>
> > HTH -- Mark D Powell --
>
>
> He should run the performance trace (event 10046). That's the first thing
>
> to do in case of a performance problem.
>
> --
>
> Mladen Gogala
>
> The Oracle Whisperer
>
> http://mgogala.byethost5.com

Mladen, as you usually do you have a point but I would want to verify that the actual plans in use are in fact really the same before running a trace on each. What I expect the trace to show is that the problem is not in the rdbms but is in the client or network.

MJM, if the plans are really the same and traces show the same or nearly the same result then a sqlnet trace might be necessary. Information on running a SQLNet trace on 11g can be found here: C:\OracDoc\Doc11gR2\E11882_01\network.112\e10835\toc.htm

If the run times vary but v$sql_plan shows the same plan then make sure the test is using the same bind or program provided constant values. If the data is skewed then the same plan can work well for one set of values but poorly for another.

HTH -- Mark D Powell -- Received on Tue Feb 04 2014 - 01:39:08 CET