[Top] [All Lists]

RE: Every new file goes into a new ag

To: "Eric Sandeen" <sandeen@xxxxxxx>
Subject: RE: Every new file goes into a new ag
From: "Bub Thomas" <thomas.bub@xxxxxxxxxxx>
Date: Thu, 12 May 2005 17:36:41 +0200
Cc: <linux-xfs@xxxxxxxxxxx>, "Braehler Uwe" <uwe.braehler@xxxxxxxxxxx>, "Lindenkreuz Morris" <morris.lindenkreuz@xxxxxxxxxxx>, "Waldschmidt Stefan" <stefan.waldschmidt@xxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
Thread-index: AcVWR+q+IgEtQ2plRNaPC35ROzEYDgAv8Ebw
Thread-topic: Every new file goes into a new ag
the tip with the inode size helped in getting the files contiguous on
However the performance of the filesystem 250 MByte/sec is much worth
compared to the raw lvm volume which is 400 MByte/sec.
A local ADIC cvfs filesystem delivers 375 MByte/sec on the same disks.
I'm trying to get you and David Chatterton into a new thread discussing
this soon, see one of my following mails.
Thanks so far.

-----Original Message-----
From: Eric Sandeen [mailto:sandeen@xxxxxxx] 
Sent: Wednesday, May 11, 2005 6:39 PM
To: Bub Thomas
Cc: linux-xfs@xxxxxxxxxxx; Braehler Uwe; Lindenkreuz Morris; Waldschmidt
Subject: Re: Every new file goes into a new ag

Bub Thomas wrote:
> Eric,
> I can't find any rotorstep within my xfs kernel codebase. Any idea how
> can get this in?

Depends on where you got your xfs code, I guess - I think it should be 
in cvs at least.

> I'm running on a 32 bit machine that's why I can't use idnode64 right?
> My machine currently shows me isize=256.
> Am I right that mkfs -t xfs -i size=512 would double the inode size?

That's right - doubling the size of the inode will double the size of 
the fs that can keep inodes under 32 bits.


> I'll give this a try tomorrow then.
> Thanks
> Thomas

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