Re: strange index performance?

From: Thomas Finneid <tfinneid(at)fcon(dot)no>
To: Scott Marlowe <scott(dot)marlowe(at)gmail(dot)com>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: strange index performance?
Date: 2009-01-25 21:54:55
Message-ID: 497CDFAF.3040706@fcon.no
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Scott Marlowe wrote:
> On Sun, Jan 25, 2009 at 1:14 AM, Thomas Finneid <tfinneid(at)fcon(dot)no> wrote:
>> Scott Marlowe wrote:
>>> So I don't think you've found the cause of your problem with the smaller
>>> index.

Ok I understand, but why dont you think the index is the problem?

>> If so, I did the test with both indexes on exactly the same db and setup.
>> And when the index uses all four ids the insert time is larger than if I
>> only use id1,3,4.
>
> I thought it was the other way around for you, that the smaller index
> was slower.

Sorry for the mistake, I meant to say the smaller index causes the
slowest insert time.

> What version pgsql are you running? My tests were on 8.3.x on a

I am running on pg 8.2.x (kubuntu x64) with 8GB ram, 8 opteron cores and
8 disks on a Areca Raid controller

> 0.12 ms per insert with all 10,000 inserted in a transaction
> 0.24 ms per insert with each insert being individual transactions
> (i.e. no begin; commt; wrapped around them all) This is on a machine
> with a 12 disk RAID-10 array under an Areca 1680ix controller with
> 512M battery backed cache. Note that the table had no other columns
> in it like yours does.

Are you sure you mean to say 0.12 ms and not 0.12 seconds? My server
also uses an Areca RAID controller (8 disk controller model), but it
does not matter how many disks are in the stripe, its still the same
performance. So, if you get that performance then I have must have set
up postgres, the OS or the RAID controller wrong. What are the most
important configurations you did to get that performance?

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message M. Edward (Ed) Borasky 2009-01-25 21:59:58 Re: postgresql 8.3 tps rate
Previous Message Scott Marlowe 2009-01-25 21:42:43 Re: [PERFORMANCE] Buying hardware