[Top] [All Lists]

Re: Consistent throughput challenge -- fragmentation?

To: stan@xxxxxxxxxxxxxxxxx
Subject: Re: Consistent throughput challenge -- fragmentation?
From: Brian Cain <brian.cain@xxxxxxxxx>
Date: Mon, 25 Feb 2013 16:38:21 -0600
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:x-received:in-reply-to:references:date:message-id :subject:from:to:cc:content-type; bh=paNqcCSshTuoWXwYWYDO6CgUqqjZId3c/01icdGKitE=; b=vRK3UJekrsBJg1WSzEXXBEbbP3XRbIYp08eT/ORPSLpbOaWpgxx/PLPaMtq9qPjZUE S8BuSDje4cNDJCtfBj/mgQafdWwKB2EvKznfomnfyxOKNcXVPjG40KCowQPUSdzoyxpG Fhe5XK+HB9KZYPhpozj4U7mQwMYt+1Ay2eVl/Uz6LtLRKtVfy7O6OPnc+klDufp4EEIr udL3+PUBrwYabTYK9owqXqnxN2/DWRhnh8Zcu/FGznHJL6CYnRoaqMw72/gibuxHGdMR XjKie4lI77HR3YP3nHXG58/n6sG9GfNOAxW5KEQK9JSKQWx4MEtRnw1folndli1vrSHP 3Blw==
In-reply-to: <512BDA2A.5050600@xxxxxxxxxxxxxxxxx>
References: <CAEWpfG_DKJt1MmWS1tARH4OmYwpSt=A-DzwKkGcD67LuR6k=Bg@xxxxxxxxxxxxxx> <512BDA2A.5050600@xxxxxxxxxxxxxxxxx>

On Mon, Feb 25, 2013 at 3:39 PM, Stan Hoeppner <stan@xxxxxxxxxxxxxxxxx> wrote:
On 2/25/2013 10:01 AM, Brian Cain wrote:
> All,
> I have been observing some odd behavior regarding write throughput to an
> XFS partition (the baseline kernel version is  I see
> consistently high write throughput (close to the performance of the raw
> block device) to the filesystem immediately after a mkfs, but after a few
> test cycles, there is sporadic poor performance.
> The test mechanism is like so:
> [mkfs.xfs <blockdev>] (no flags/options, xfsprogs ver 3.1.1-0.1.36)
> ...
> 1. remove a previous test cycle's directory
> 2. create a new directory
> 3. open/write/close a small file (4kb) in this directory
> 4. open/read/close this same small file (by the local NFS server)
> 5. open[O_DIRECT]/write/write/write/.../close a large file (anywhere from
> ~100MB to 200GB)

It looks like omitting step #2 and putting all files in the root directory (and modifying step #1 to remove only the files that were associated with that previous test cycle) avoids the poor-performance problem as well.  At least, that's the case in testing so far.

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