NetBSD-Users archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: [netbsd-6] Status of SSD TRIM support

On Thu, Nov 08, 2012 at 06:08:32PM +0000, Hugo Silva wrote:
> On 11/08/12 14:25, Petar Bogdanovic wrote:
> > On Thu, Nov 08, 2012 at 12:54:45PM +0000, Hugo Silva wrote:
> >>
> >> Just wondering if NetBSD 6 is currently supporting TRIM on SSDs.
> > 
> > I remember reading some recent articles questioning the benefits of TRIM
> > in conjunction with newer SSDs.  The reasoning behind the articles was
> > that the current TRIM command can't be queued so small deletes may lower
> > the overall performance and that modern SSD controllers don't need TRIM
> > anyway.
> > 
> Interesting. Do you still have the articles mentioned at hand?

It's possible that some of these articles were lengthy lkml mails.  Here
is one article I found through Google:

        by Theodore Ts'o, Ext4 Maintainer


        That being said, even if you don't use TRIM, and even if you
        fill the disk to 100% of its stated capacity, the SSD will
        always have some "elbow room", because it is designed that way.
        The result of not using TRIM and filling the disk to 100% of
        capcity is that the SSD will probably get a little slower,
        especially in random write workloads.  How much slower will
        depend on the particular SSD's controller, and how much spare
        capacity was designed into the drive.


It's probably worth noting that performance degradation was real and
reproducible back in 2008 on some of the first Intel SSDs (X25).  Since
then, though, there have been several benchmarks where you could not
tell the difference between TRIM and no TRIM, especially when recent
Sandforce controllers were involved (unfortunately the ones that later
got a bad reputation for having high failure rates).

Sorry for the lack of sources.

Home | Main Index | Thread Index | Old Index