[Top] [All Lists]

Re: XFS, RAID-5, 3ware

To: Gaspar Bakos <gbakos@xxxxxxxxxxxxxxx>
Subject: Re: XFS, RAID-5, 3ware
From: Chris Wedgwood <cw@xxxxxxxx>
Date: Mon, 23 May 2005 20:08:07 -0700
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <Pine.SOL.4.58.0505232234410.2452@xxxxxxxxxxxxxxxxxxxxx>
References: <Pine.SOL.4.58.0505191735250.26773@xxxxxxxxxxxxxxxxxxxxx> <c4cbdce7bad01d946ce6fa0c57872567.ANY@xxxxxxxxxxxxxxxxxxxxx> <Pine.SOL.4.58.0505232234410.2452@xxxxxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
On Mon, May 23, 2005 at 10:46:13PM -0400, Gaspar Bakos wrote:

> Here you refer to the kernel page-size?


> > 512-bytes is the minimum you can get away with to spread inodes out
> > evenly over a volume larger than 2TB
> Ok, I see.

i should qualify that though that for >4TB you will need larger inodes
again but at that point i think using a 32-bit CPU won't be very

> What tests do you think of?

whatever matches you usage pattern, this is why i mentioned that the
strip-size doesn't necessarily have some *magic* value --- it will
vary somewhat depending on what your writes look like

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