xfs
[Top] [All Lists]

Re: Strange system behaviour when copying disks

To: linux-xfs@xxxxxxxxxxx
Subject: Re: Strange system behaviour when copying disks
From: Ethan Benson <erbenson@xxxxxxxxxx>
Date: Thu, 15 May 2003 22:25:00 -0800
In-reply-to: <1053005705.952.13.camel@venus>
Mail-copies-to: nobody
Mail-followup-to: linux-xfs@xxxxxxxxxxx
References: <1052916810.2395.15.camel@venus> <1053005705.952.13.camel@venus>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.3.28i
On Thu, May 15, 2003 at 03:35:05PM +0200, Olaf Fr?czyk wrote:
> 
> With vanilla 2.4.20 it is the same as 2.4.20 with XFS. Load is about 5,
> CPU is about 90% idle. 'cat' takes about 3-5 % CPU.
> 
> What is really strange, if I use 'dd' eg.:
> dd if=/dev/hda1 of=/dev/hdc1 bsize=1024 count=1000000
> there are NO freezes at all.
> The system works smoothly. Load is about 1.2.
> dd takes for some time 30% CPU and then until finishes about 0.2 % CPU.
> This is for radhat, vanilla and vanilla+XFS kernels.
> 
> So it is not caused by XFS code. I'll take it to LKML.
> 
> BTW, does somebody know why there is such strange difference between
> doing 'dd' and 'cat'?
> 
> The copy speed is the same for 'dd' and 'cat': about 6.6 MB/sec

what happens if you use bs=4096 to dd ? 

doing a strace of cat shows it appears to do read/writes in 4096 byte
blocks perhaps this is making a difference.

-- 
Ethan Benson
http://www.alaska.net/~erbenson/

Attachment: pgp3lR11cNmuD.pgp
Description: PGP signature

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