[Top] [All Lists]

trim performance on xfs (also reply to: Online TRIM/discard performance

To: xfs@xxxxxxxxxxx
Subject: trim performance on xfs (also reply to: Online TRIM/discard performance impact)
From: Andrei Purdea <andrei@xxxxxxxxx>
Date: Mon, 12 Dec 2011 00:17:17 +0200
Cc: martin.rusko@xxxxxxxxx
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=purdea.ro; s=google; h=mime-version:x-originating-ip:from:date:message-id:subject:to:cc :content-type; bh=6UXCXhB0p7tSZ/x6cUI9Y7VcTBfvk7mfLHhIyIKaEns=; b=cYzQH3nKvFT7CbQ4U06CPK3kkuDy/kkYG1EqX0Q6xa+qr+hl1I+gKhCGln6enyrXLF of0BQSL10K7A1sP4z5b6rLw5jG3uihll10WaTJZ33Tw5FCpezoIAyHLBfMCOsfTKfrIQ E6cm9ZXZxqYvyJGMaEJPimFG9XYlEWNS0z0KE=
I also recently met with an issue that has been described in a previous email

I have an OCZ Agility 3 (also a SandForce 2281 controller),
and a Corsair Performance Pro (non-SandForce controller,
with good non-compressible data speeds)

Removing a linux-3.2-rc5 tree took more then 6 minutes with online discard,
and ext4 was very speedy (less then a second)
, but only on the SandForce controller!!
For the Corsair performance was good on both filesystems.

So I decided to debug what is happening.
I also noticed that fstrim was slower on the OCZ
(30-40 seconds, vs. 3 seconds on the Corsair)
So I had a hunch that the performance difference comes from the number of
sectors that are discarded in a single TRIM command.

So i compiled a kernel where I inserted a printk into blkdev_issue_discard(),
to see how many sectors each call asks for.


Total number of sectors discarded:
    ext4: 1032864   XFS: 1043112 (about the same)
Number of calls to blkdev_issue_discard():
    ext4: 24            XFS: 39395
Average number of sectors discarded per call:
    ext4: 43036 (21 MiB)   XFS: 26.48  (13.2 KiB)

I made some TRIM performance measurements (see ssdperf.png attached)
with a small tool i wrote. (warning, destructive to data)

It looks like the OCZ needs a really high amount of time to execute
even the smallest TRIM command.

I made a small post about this here too: http://purdea.ro/projects/trim_perf/

Can this be fixed in XFS?

I am curious, do all SandForce controllers exhibit this slow TRIM issue?


Purdea Andrei
student at the “Politehnica” University of Timisoara
Faculty of Automation and Computers (AC)
Master's Degree Program - Computer Engineering (MCE)
2nd year

Attachment: xfs.txt
Description: Text document

Attachment: ext4.txt
Description: Text document

Attachment: ssdperf.png
Description: PNG image

Attachment: trim_perf.c
Description: Text Data

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