Re: Really bad insert performance: what did I do wrong?

On Fri, Feb 21, 2003 at 10:32:37AM -0500, Kevin White wrote:
> I've use PostgreSQL for some pretty amazing things, and was a proponent
> of using it here. I set up 7.3.1 on my development RedHat 8 box, and it
> was fine. I need to load about 700,000 rows into one table, the rows
> only having 6 columns, and the load on my box happens in just a couple
> of minutes (there's some calculation while the data is loading, and that
> time was acceptable to me).
>
> My box, however, isn't a production server, so I attempted to create the
> database again on a Sun Blade:
>
> SunOS trident 5.8 Generic_108528-17 sun4u sparc SUNW,UltraAX-i2
> Status of processor 0 as of: 02/21/03 10:10:10
> Processor has been on-line since 01/13/03 13:53:51.
> The sparcv9 processor operates at 500 MHz,
> and has a sparcv9 floating point processor.
>
> It isn't the world's fastest box, but should be fast enough for this.

What's the disk subsystem? Is fsync turned on in both cases? And is
your IDE drive lying to you about what it's doing.

My experiences in moving from a Linux box to a low-end Sun is pretty
similar. The problem usually turns out to be a combination of
overhead on fsync (which shows up as processor load instead of i/o,
oddly); and memory contention, especially in case there are too-large
numbers of shared buffers (on a 16 Gig box, we find that 2 Gig of
shared buffers is too large -- the shared memory management is
crummy).

A

--
----
Andrew Sullivan 204-4141 Yonge Street
Liberty RMS Toronto, Ontario Canada
<andrew(at)libertyrms(dot)info> M2P 2A8
+1 416 646 3304 x110