Re: Reliability recommendations

From: "Luke Lonergan" <llonergan(at)greenplum(dot)com>
To: "Bruce Momjian" <pgman(at)candle(dot)pha(dot)pa(dot)us>, "Christopher Browne" <cbbrowne(at)acm(dot)org>
Cc: pgsql-performance(at)postgresql(dot)org
Subject: Re: Reliability recommendations
Date: 2006-02-24 15:14:20
Message-ID: C02462CC.1DB8A%llonergan@greenplum.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Bruce,

On 2/24/06 6:29 AM, "Bruce Momjian" <pgman(at)candle(dot)pha(dot)pa(dot)us> wrote:

> Christopher Browne wrote:
>> After takin a swig o' Arrakan spice grog, jd(at)commandprompt(dot)com ("Joshua D.
>> Drake") belched out:

Always more fun to read drunken posts :-)

>>>> Dell 2850
>>>> 2 x 3.0 Ghz Xeon 800Mhz FSB 2MB Cache
>>>> 4 GB DDR2 400 Mhz
>>>> 2 x 73 GB 10K SCSI RAID 1 (for xlog and OS)
>>>> 4 x 146 GB 10K SCSI RAID 10 (for postgres data)
>>>> Perc4ei controller

> Dell often says part X is included, but part X is not the exact same as
> part X sold by the original manufacturer. To hit a specific price
> point, Dell is willing to strip thing out of commodity hardware, and
> often does so even when performance suffers. For many people, this is
> unacceptable.

I'll register the contrarian's point of view on this: we just had a customer
(an airline) buy the exact machines that Jeremy lists here, including the
Perc4 controller (an LSI MPT RAID controller).

Besides the complete lack of real RAID 10 support, the machines are actually
performing very well in hardware RAID5 mode. They are running Redhat 4
linux, and we did have to tune the I/O readahead to get the performance we
needed - about 250MB/s on 2 banks of RAID5, not bad at all for only 4 active
disks.

I think the build quality and the fit/finish/features of their chassis were
all pretty good.

So, if you want RAID5, these machines work for me. The lack of RAID 10
could knock them out of contention for people.

The problem with their RAID10 is actually hidden from view BTW. You can
configure the controller with a "spanning" and other options that sound like
RAID10, but in actuality it's spanning pairs of RAID1 disks, which does not
provide the performance benefit of RAID10 (see James Thornton's post here:
http://openacs.org/forums/message-view?message_id=178447), the important bit
of which says:

"RAID-10 on PERC 2/SC, 2/DC, 3/SC, 3/DCL, 3/DC, 3/QC, 4/Di, and CERC
ATA100/4ch controllers is implemented as RAID Level 1-Concatenated. RAID-1
Concatenated is a RAID-1 array that spans across more than a single pair of
array disks. This combines the advantages of concatenation with the
redundancy of RAID-1. No striping is involved in this RAID type. Also,
RAID-1 Concatenated can be implemented on hardware that supports only RAID-1
by creating multiple RAID-1 virtual disks, upgrading the virtual disks to
dynamic disks, and then using spanning to concatenate all of the RAID-1
virtual disks into one large dynamic volume. In a concatenation (spanned
volume), when an array disk in a concatenated or spanned volume fails, the
entire volume becomes unavailable."

- Luke

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Luke Lonergan 2006-02-24 15:41:12 Re: Reliability recommendations
Previous Message Craig A. James 2006-02-24 15:04:53 Re: Reliability recommendations