[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On Thu, 29 Apr 2010, John Williams wrote:
On Thu, 2010-04-29 at 07:40 +0100, tom wrote:I think things like swap can still cause problems so you can still wipe one out a lot faster than 5 years.Yea I have no /swap at all, I have a 2GB partition on an HDD that I can just point it at if the need arises. I've also move /tmp into RAM with the tmpfs doodah. I don't intend having the ssd very full at all, leaving it room to shift stuff around wear levelling. I didn't leave the suggested 10% or 20% unpartitioned that I have seen some suggest. I *think* by just not filling the drive it has the same effect but taking will power and organisation rather than an enforced lower capacity :]
I'd be curious to know how to do it - the write wear leveling that is. Unless they have intimate knowledge of the filesystem they can only do a one-of logical block mapping to sectors the SSD knows have never been written to before - once that sector has been written, the SSD won't know if it's free or not.
And thinking about it now, I do a full badblocks write/read test of all flash drives I install in my units - so that's probably rendered their wear leveling useless. (but I don't think CF and the flash IDE drives I use have wear leveling anyway)
Hmm... http://en.wikipedia.org/wiki/Wear_levelling Gordon -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/linux_adm/list-faq.html