On 14/02/11 22:38, Simon Royal wrote:
Ive been reading about the limited writes on SSD, with conflicting info.
Does anyone have a clear info on this. Do they have limited write capabilities? Is this within a normal users/machines lifespan? What happens when it reaches it write limit.
Short answer is that yes they do have a limited (write) life although unless you do something intensive that is always going to write to the same blocks it is unlikely you will hit it in normal lifespan given reasonable quality and modern flash ram.
There are tricks to reduce write such as use a non-journalised file system like ext2 instead of ext3/4 and do not use a swap partition.
Yes that helps and everything on (cheaper) flash that avoids writes will also give you a speed boost as writes generally take more time (and power) than reads.
So in addition to that you might want to mount with the noatime option unless you really need access timestamps (unlikely)
Maybe you could also mount /var/log and possibly /var/tmp as tmpfs if it is the sort of machine where you aren't too worried about keeping log files between reboots.
I would recommend either not using a swap file (assuming you have lots of ram and don't care about suspend to disk) or if you do maybe look at the vm.swappiness kernel parameter to reduce how much it is used.
Finally I have also seen boosts by changing from the default cfq (on most distro's) to the noop I/O scheduler in much the same way as I have seen on VM's but YMMV
I really hope someone doesn't repost this question under a different Subject after I have taken all this effort to reply once already...that would be annoying.