xfs
[Top] [All Lists]

Re: suppressing syncs on spools

To: Bernhard Erdmann <be@xxxxxxxxxxx>
Subject: Re: suppressing syncs on spools
From: Jan-Frode Myklebust <janfrode@xxxxxxxxxxx>
Date: Sat, 8 Mar 2003 14:11:10 +0100
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <1047052002.3e68bee241784@xxxxxxxxxxxxxxxxxx>; from be@xxxxxxxxxxx on Fri, Mar 07, 2003 at 04:46:42PM +0100
References: <1047052002.3e68bee241784@xxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Fri, Mar 07, 2003 at 04:46:42PM +0100, Bernhard Erdmann wrote:
> 
> for a spool/scratch filesystem I'm only interested in speed. Disk I/O
> should be as minimal as possible. Files will be deleted some 1/10 sec or
> some secs later. It won't hurt if files are lost if the box crashes.
> This filesystem can even be mkfs'd at system boot.
> 
> Just speed, like a ram disk, but I don't want to spend 1-2 GB to a ram disk.

Have you considered tmpfs? tmpfs can live in both ram and disk (swap).
Read linux/Documentation/filesystems/tmpfs.txt.

> 
> Are there mount options or mkfs.xfs options which can help to reduce
> disk I/O?

Don't know of any, except maybe putting the log on a ramdisk? And of
course 'noatime'.


  -jf


<Prev in Thread] Current Thread [Next in Thread>