Re: Oracle Myth II - ??

Chuck wrote:
>
> My initial tests proved pretty much what I thought. I copied a 1,000,000 row
> 300m table across a database link with a single insert statement. It took 97
> seconds.
>
> After truncating the destination table, I ran two inserts simultaneously
> selecting from the same source and inserting into the same destination. The
> first insert copied all of the even numbered IDs while the second one did
> the odd numbered ID's. It took 65 seconds.
>
> The platform used was Oracle 8.1.7 on AIX 4.3 on both ends of the link. The
> two databases were on two different servers connected by TCP/IP over a 100mb
> ethernet.
> --
> Chuck
>
> "Chuck" <chuckh_at_softhome.net> wrote in message
> news:alaq5q$1mvtdu$1_at_ID-85580.news.dfncis.de...
> > Without testing it, off the top of my head I would say yes, there probably
> > is a performance benefit. With any type of i/o whether it be disk or
 network
> > based, there is some time spent in overhead. That time could be used for
> > parallel i/o processing to take place.
> >
> > I will do a small test on a table with a few hundred meg and post results
 if
> > anyone is interested.
> > --
> > Chuck
> >
> > "Ed Stevens" <spamdump_at_nospam.noway.nohow> wrote in message
> > news:3d78bd02.64028267_at_ausnews.austin.ibm.com...
> > > Some time ago we had a very interesting and informative thread on Oracle
 myths,
> > > mostly centering on tablespace fragmentation and number of extents as a
 function
> > > of performance.
> > >
> > > I have a question about another possible myth, though I've never seen
 any
> > > discussion on the subject, either pro or con.
> > >
> > > The question is: If one needs to move large quantities of data through
 a
 db
> > > link, is there any performance advantage (supposedly from
 parallelization)
 of
> > > splitting the work through 2 separate links? Do 2 links represent a
 2-lane
> > > road, or simply two different ramps onto a single lane road? If there
 would be
> > > a performance gain, how would an application actually implement it?
> > >
> > > Platform is Oracle EE 7.1.3 on Win2K. The two databases involved would
 be
 on
> > > different boxes.
> > > --
> > > Ed Stevens
> > > (Opinions expressed do not necessarily represent those of my employer.)
> >
> >

Maybe you could repeat the exercise and post the results of v$session_event et al - it would be interesting to see where the time was lost

Cheers
Connor

-- 
==============================
Connor McDonald

http://www.oracledba.co.uk

"Some days you're the pigeon, some days you're the statue..."