Re: SQL*Net message from client waits

On 28 Jun., 12:22, "Jonathan Lewis" <jonat..._at_jlcomp.demon.co.uk> wrote:
> <baba_j..._at_tahiti.net> wrote in message
>
> news:1182973560.318722.20190_at_k29g2000hsd.googlegroups.com...
>
> > Hello
>
> > some users complain about a Powercenter performance, I traced the
> > sessions for an hour, please have a look at last section of the the
> > output of tkprof below. It is a 32 processors Sun running 9.2.0.6, not
> > dedicated to this application. Running the trace the box was in avg.
> > 30% idle. I checked the plans and they seem to be OK, I know that
> > these wait events *may* be idle. Do I have problems with the network
> > in your opinion? Has anybody dealt with comparable problems?
>
> For comments, see
> http://jonathanlewis.wordpress.com/2007/06/28/tkprof-1/

Reading this I was taken aback, one of the best short stories I read in the last time. Chapeau bas, the long time to go for myself, I realised. I liked and like your way of thinking, it was this breath of fresh air.

Coming down to earth: sorry for the crude statement of the facts, you understand why. I meant in fact not a network perf, rather communication problems. It is a 3-tier environment, the goal of this exercise is, as usually, to deliver a prove of Oracle´s innocence, what is somehow for me obvious. At least sometimes.

As you suggested, I jumped today into the trivial and banal reality trying to sort out, step by step, the worst single operations, reducing their logical io, having a look on existing indexes and so on. Btw. with some quite promising results. After that I'll try to rerun the trace and *maybe* jump to conclusions.

In any case: many thanks for your time, very inspirational. But frustrating as well :-)

Regards
Ch. Received on Thu Jun 28 2007 - 16:02:57 CDT