Re: Moving pg_xlog

From: John A Meinel <john(at)arbash-meinel(dot)com>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: KeithW(at)narrowpathinc(dot)com, PostgreSQL Perform <pgsql-performance(at)postgresql(dot)org>
Subject: Re: Moving pg_xlog
Date: 2005-06-01 21:27:48
Message-ID: 429E2854.1050209@arbash-meinel.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-performance

Tom Lane wrote:
...

>Now that I think about it, you were (if I understood your layout
>correctly) proposing to put the xlog on your system's root disk.
>This is probably a bad idea for performance, because there will always
>be other traffic to the root disk. What you are really trying to
>accomplish is to make sure the xlog is on a disk spindle that has no
>other traffic besides xlog, so that the disk heads never have to move
>off the current xlog file. The xlog traffic is 100% sequential writes
>and so you can cut the seeks involved to near nil if you can dedicate
>a spindle to it.
>
>
I certainly agree with what you wrote. But my understanding is that if
you only have 2 arrays, then moving xlog onto the array not on the
database is better than having it with the database. It isn't optimum,
but it is better. Because that way there isn't as much contention
between the database and xlog.

John
=:->

In response to

Responses

Browse pgsql-performance by date

  From Date Subject
Next Message Tobias Brox 2005-06-02 02:05:28 Forcing use of specific index
Previous Message Brad Nicholson 2005-06-01 20:31:38 Re: Moving pg_xlog