Re: Does (block)Size really matter?

On 6/8/05, Smith, Ron L. <rlsmith_at_kmg.com> wrote:
> I have a process that sends data through a dblink to an intermediate
> database and then on to another database. The first database is blocked
> at 4096, the second is at 8192, the third at 8192. Response is slow
> (5000 rows in 1.5 hours. Could the difference in block size be a
> factor?

>=20

Ron,

Check out the wait events in that session - along the lines of 'SQL*Net' and 'db_link'.
Perhaps hiking the SDU/TDU in the client tnsnames.ora and server listener.ora (use a separate listener) to 8192, 16384 or some multiple of 42 could reduce the number of round trips.

see Metalink for references, e.g.=20
http://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_data= base_id=3DNOT&p_id=3D44694.1

There was lots of run with this in 9.2 where dynamic resgistration would not support non-standard SDU sizes.

At least you can setup separate connect descriptors and listeners, seeing that this is only for server to server communications - its not like you have to hit all the desktops. Might be worth a shot.

hth.

Paul

--
http://www.freelists.org/webpage/oracle-l