Why is SU+J undesirable on SSDs?

Jeff Roberson jroberson at jroberson.net
Sat Nov 3 23:16:04 UTC 2012

On Sat, 3 Nov 2012, Ian Lepore wrote:

> On Sat, 2012-11-03 at 17:06 -0500, Adam Vande More wrote:
>> On Sat, Nov 3, 2012 at 4:30 PM, Brett Glass <brett at lariat.net> wrote:
>>> Have been following the thread related to SU+J, and am wondering: why is it
>>> considered to be undesirable on SSDs (assuming that they have good wear
>>> leveling)?
>> Superstition
> Yeah, that's what it must be.  Or... it could be well-informed choice.
> Journaling increases the number of writes.  That puts wear on any disk,
> mechanical or SSD, and it takes time.  What it buys you is better
> performance if you get into a crash recovery situation.  It's perfectly
> reasonable for someone to make the decision that their SSD can finish an
> fsck so fast that there's no point in paying any penalty for the extra
> writes for journaling.

The journal entries are 32 bytes per in SUJ.  So the number of extra 
writes is down in the noise.  The journaling also gets you asynchronous 
partial truncation and a few other asynchronous operations that are sync 
in SU.  It does cost slightly more cpu time and more memory.  I'm not 
saying you're making the wrong choice.  I'm just saying that it's not 
clear that you should or should not use it with SSDs.

> I have a 256G SSD here with about 200G of data on it, and fsck without
> journaling takes about 3 minutes.  I can live with that.  With more data
> or a slower drive I might make a different choice.

If you are happy with 3 minutes this is very reasonable and I assume you 
turn off bg fsck.


> -- Ian
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at freebsd.org"

More information about the freebsd-stable mailing list