Message-ID: <9473692.post@talk.nabble.com>
Date: Wed, 14 Mar 2007 05:46:55 -0700 (PDT)
From: femski <hypertree@yahoo.com>
To: pgsql-performance@postgresql.org
Subject: Re: Postgres batch write very slow - what to do
In-Reply-To: <9453712.post@talk.nabble.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
References: <9452092.post@talk.nabble.com> <45F685B7.5040001@enterprisedb.com>
	<9453712.post@talk.nabble.com>


Ok, I turned off XDMCP and network bandwidth utilization dropped to less than
5%.
Timings remained the same.

Curiously five times faster time for Oracle came from a client running on a
different host than the server.
To make things worse for Postgres, when I replace "hostname" in jdbc string
to "localhost" or 127.0.0.1
it runs another 60% slower (446 sec vs 275 sec). Strange.

Before I take this discussion to jdbc list, why is CPU utilization 100%
during insert ? could that be a bottleneck. How to eliminate it ? These are
Intel WordCrest  5110 Xeon cores.

thank you

-Sanjay


femski wrote:
> 
> I am runing Postgres 8.2 on OpenSuse 10.2 with latest jdbc driver. I moved
> the app to be collocated with the server. Oracle takes 60 sec. Postgres
> 275 sec. For 4.7 million rows.
> 
> There are 4 CPUs on the server and one is runing close to 100% during
> inserts.
> Network history shows spikes of upto 60% of the bandwidth (Gnome System
> monitor graph). I have a gigabit card - but should not enter into picture
> since its on local host.
> 
> thanks,
> 
> -Sanjay
> 

-- 
View this message in context: http://www.nabble.com/Postgres-batch-write-very-slow---what-to-do-tf3395195.html#a9473692
Sent from the PostgreSQL - performance mailing list archive at Nabble.com.