Re: Ideal hardware configuration for pgsql

From: Bruce Momjian <pgman(at)candle(dot)pha(dot)pa(dot)us>
To: Gordan Bobic <gordan(at)freeuk(dot)com>
Cc: pgsql-general(at)postgresql(dot)org
Subject: Re: Ideal hardware configuration for pgsql
Date: 2001-05-04 16:29:05
Message-ID: 200105041629.f44GT5X24041@candle.pha.pa.us
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-general

> > > A note about SCSI vs IDE... I have recently tried both on a dual
> P3 with
> > > 1gb of ram running Mandrake 7.2. I was amazed the idle CPU's were
> > > running near 20-23% with nothing other then a bash shell running
> on 2
> > > IBM IDE ATA 100 drives. I converted to 2 IBM SCSI U2 drives and
> the idle
> > > CPU's went down to 0-2%. This was a tremendous difference. I feel
> SCSI
> > > is a must.
>
> I have a suspicion that you didn't try using hdparm first to tune up
> your disks. Have you set them up with 32-bit transfers, UDMA, and
> unmasked interrupts? I am running a single 40 GB Maxtor 5400 UDMA/66
> disk, and even when I am stress testing the database, about 96% on
> each CPU goes to user-space postmaster processes. The system CPU
> consumption rarely goes above 4%. And FYI, I am running a dual 1 GHz
> P3 on a VP6. The IDE disk is connected to thepromise controller. I
> haven't tried it on the VIA, but I suspect the results would be the
> same. The only way I can see your CPU consumption going to 20% is if
> you are using PIO, and haven't tuned up your disks.

I will ditto that. IDE is terrible _unless_ you are using DMA for data
transfer. Without DMA, the CPU is involved in transfering all data
to/from the drive, hence your 20% CPU utiliation and terrible
performance.

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026

In response to

Browse pgsql-general by date

  From Date Subject
Next Message Joel Burton 2001-05-04 16:36:18 Re: Metaphone function attachment
Previous Message Stephan Szabo 2001-05-04 16:16:12 Re: a primer on trigger?