Date: Fri, 28 Aug 2015 13:51:02 +0100 From: RW <rwmaillists@googlemail.com> To: freebsd-questions@freebsd.org Subject: Re: Replacing Drive with SSD Message-ID: <20150828135102.79c52f02@gumby.homeunix.com> In-Reply-To: <55E0266B.10005@infracaninophile.co.uk> References: <CEAD84AD-341A-4FB9-A3A1-D0D5A550AFFD@lafn.org> <55E01DAE.1020709@infracaninophile.co.uk> <20150828084643.GB1274@xtaz.uk> <55E0266B.10005@infracaninophile.co.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 28 Aug 2015 10:14:19 +0100 Matthew Seaman wrote: > On 08/28/15 09:46, Matt Smith wrote: > > I've heard a rumour that you should never use dd with SSD drives > > because of the wear levelling stuff. Apparently SSDs automatically > > make sure that data is sent to unused flash cells so that all the > > cells wear evenly, but if you use dd on them it makes them think > > that every single cell is in use which screws this up? > > Hmmm.... Yes, dd will copy all of the source disk including disk > blocks that are unused, empty space. Overwriting a cell that is > already zeroes with yet more zeroes is a waste of time, They wont necessarily be zeros. > but I don't > know if that would actually use up some of the life of that cell. It > shouldn't confuse the wear-levelling code on the drive particularly > -- it might take a little while to sort itself out after the fact, The problem is that if you write to the whole device you reduce the free blocks to the over-provisioning level. Whether or not that's a problem depends on whether the device has static wear-levelling and how good it is. Without it the writes all go into a relatively small pool of blocks. When I bought my SSD last year I couldn't see any evidence that consumer grade SSDs have static wear-levelling. I think it would be mentioned if they did, as there's so much online about working around its absence by leaving a large free block pool.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150828135102.79c52f02>