Re: [GENERAL] Arguments Pro/Contra Software Raid - Mailing list pgsql-performance

From Bruce Momjian
Subject Re: [GENERAL] Arguments Pro/Contra Software Raid
Date
Msg-id 200605101435.k4AEZxc13602@candle.pha.pa.us
Whole thread Raw
In response to Re: [GENERAL] Arguments Pro/Contra Software Raid  (Vivek Khera <vivek@khera.org>)
List pgsql-performance
Vivek Khera wrote:
>
> On May 10, 2006, at 12:41 AM, Greg Stark wrote:
>
> > Well, dollar for dollar you would get the best performance from
> > slower drives
> > anyways since it would give you more spindles. 15kRPM drives are
> > *expensive*.
>
> Personally, I don't care that much for "dollar for dollar" I just
> need performance.  If it is within a factor of 2 or 3 in price then
> I'll go for absolute performance over "bang for the buck".

That is really the issue.  You can buy lots of consumer-grade stuff and
work just fine if your performance/reliability tolerance is high enough.

However, don't fool yourself that consumer and server-grade hardware is
internally the same, or has the same testing.

I just had a Toshiba laptop drive replaced last week (new, not
refurbished), only to have it fail this week.  Obviously there isn't
sufficient burn-in done by Toshiba, and I don't fault them because it is
a consumer laptop --- it fails, they replace it.  For servers, the
downtime usually can't be tolerated, while consumers usually can
tolerate significant downtime.

I have always purchased server-grade hardware for my home server, and I
think I have had one day of hardware downtime in the past ten years.
Consumer hardware just couldn't do that.

As one data point, most consumer-grade IDE drives are designed to be run
only 8 hours a day.  The engineering doesn't anticipate 24-hour
operation, and that trade-off passes all the way through the selection
of componients for the drive, which generates sigificant cost savings.

--
  Bruce Momjian   http://candle.pha.pa.us
  EnterpriseDB    http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +

pgsql-performance by date:

Previous
From: Wu Fengguang
Date:
Subject: Re: PostgreSQL VACCUM killing CPU
Next
From: PFC
Date:
Subject: Re: [HACKERS] Big IN() clauses etc : feature proposal