Re: Slow database performance... possible SQL*NET problem?

From: David Ng <david_ng_at_mindlink.bc.ca>
Date: 1996/11/23
Message-ID: <3296BD87.116D_at_mindlink.bc.ca>#1/1


Phil,

If SQL*Net causes problems in import (I assume that is Oracle import utility), may be you should consider doing the import in HP-UX. In other word, you eliminate the SQL*Net in this import process.

David Ng

Phil M Badger wrote:
>
> Hi,
>
> I'm posting this message to both the Oracle and HP-UX newsgroups in the
> hope that someone can assist with a problem we are experiencing with poor
> database performance.
>
> The installation in question is SAP R/3 v3.0D, running on top of an Oracle
> 7.2.3 database and HP-UX 10.10 (all housed on a HP 3000 K410 system). I am
> in the process of setting up this system and one of the tasks required is a
> foreign language import and this is where I'm seeing the performance hit.
> I had to perform the same task on other systems of near identical
> configuration and the job ran for about 3 - 3.5 hours -- all it's doing is
> inserting records into a selection of tables. However, on the latest
> system, it's been running for about 24 hours already and is nowhere near
> completion. I can monitor the import through R/3 and the task is
> processing about 4 records per second -- on the previous installations, it
> went through about 150 recs/sec.
>
> In monitoring the system whilst this job runs, I noted that I/O, CPU and
> memory activity were all low (CPU usage doesn't get much above 3%), but
> network activity was going crazy -- we're talking about 700+ packets in/out
> per second.
>
> My guess as to what's causing the problem is SQL*NET, which seems to think
> that it's communicating with a remote database. However, I'm no DBA and
> I'm at a bit of a loss at what to do to modify this behaviour (SAP don't
> see fit to ship Oracle documentation with their product). Can anybody
> advise?
>
> Just a little more background information -- the system is an NIS client
> (as was the other system on which the import ran smoothly) on the same
> segment as the master server. Also, output from the netstat command for
> the offending server shows about 20 active connections from hostname.tlisrv
> back to itself (hostname.socket #). More proof that SQL*NET's not set up
> correctly.
>
> I would be grateful for any suggestions on what I need to do to remedy this
> matter.
>
> Thanks,
> Phil Badger
> pbadger_at_cultorfs.com
  Received on Sat Nov 23 1996 - 00:00:00 CET

Original text of this message