Re: Re: Fast Inserts and Hardware Questions - Mailing list pgsql-general

From Steve Wolfe
Subject Re: Re: Fast Inserts and Hardware Questions
Date
Msg-id 000d01c0acee$3b1b1b40$50824e40@iboats.com
Whole thread Raw
In response to Fast Inserts and Hardware Questions  (Orion Henry <orion@trustcommerce.com>)
List pgsql-general
> > > * Is there an architecture best suited for this.  Should I shell out
the
> > > extra cash for a 64 bit box over a 32bit one.  Quad Xeon or Quad
Alpha?
> > > Quad Alpha or Quad UltraSparc?
> >
> > Again heresay, but Sun's hardware has excellent I/O throughput and it's
> > difficult to get people to abandon it in favour of Intel once they've
> > tried it.
>
> What I was hoping to hear was something along the line of "An int8
> index?  Dear God!  You better be using a 64 bit machine!"  or  "The only
> way you will be able to handle a load like that will be with
> UltraSparc+Solaris!!" or "Some Alpha fan did tons of assembly level
> optimizations for the Alpha processors, they are the fastest!"  From the
> sounds of it what I need is lots of fast disks and to hell with the
> CPU.  I might as well get an Athlon and a Network Appliance.

     I have a Quad Xeon that I could do some testing for you, if you wanted.
I'm awfully lazy, though, so you'll have to send me a schema and some test
data. ; )  I can give it a run with a single process, or with multiple
simultaneous processes over the network.

steve

    Alphas may be terrific (amazing, actually) for floating-point work, but
for Postgres work, I'm not terribly impressed with them.  A Compaq rep let
us borrow a $25,000 dual-alpha for a week, and I pitted in in a severe
stress-test against our quad Xeon.  At the time, because of a lack of money
and an obscure bug, we had PII Xeon 400's in it, *without* the L2 cache -
talk about cutting the hamstrings.  The Compaq machine was only about 1.5
times faster than the Xeon.  Now that we have PIII Xeon700's in it (with the
cache working), the machine has probably run us about $12,000 (including
hardware RAID and the lot), and I'm sure that it would at least equal the
dual Alpha now.

steve



pgsql-general by date:

Previous
From: Antoine Reid
Date:
Subject: Re: Sluggish INSERTS with Foreign Keys (7.1beta5)
Next
From: Tom Lane
Date:
Subject: Re: Sluggish INSERTS with Foreign Keys (7.1beta5)