[ltp] Re: [fstab: noatime?]

Paul Seelig linux-thinkpad@linux-thinkpad.org
Sat, 15 Nov 2014 21:47:22 +0100


On 11/15/2014 07:47 PM, Uwe Brauer wrote:
>  nodiratime and noatime should prolong the life expectation of the SSD
> without much performance loss. 
> 
All the contrary, it will even enhance performance to some degree!

I used this already with HDD's because i never saw any real advantage in
extra disk accesses just for updating the inode with superfluous
information nobody and nothing has any need for anyway. In my opinion,
not using both nodiratime and noatime just slows down disk performance
for basically no gain at all.

Furthermore, unless you bought one of those cheaper TLC based SSD's, i
really wouldn't bother about any futile measures to extent the life span
of a SSD. The idea that SSD's have a severely limited life time is a
myth at best, most probably due to the confusion between static flash
based memory gadgets (USB sticks, etc.) with SSD's, which unlike the
former are featuring sophisticated controllers and strategies to avoid
excessive wear.

http://en.wikipedia.org/wiki/Solid-state_drive#Comparison_with_memory_cards

http://en.wikipedia.org/wiki/Solid-state_drive#Wear_leveling

For the very same reason i also wouldn't want to renounce having a swap
partition on SSD. Besides the faster file access and write during daily
work, taking advantage of the much faster suspend and wake up from
hibernation is also one of the main reasons to use an SSD.

MLC based SSD's will most probably not be worn down just by using it.
Even if continuously writing to and deleting data from it will not do
much harm during the course of various years.

Some recommended reading:

http://en.wikipedia.org/wiki/Solid-state_drive

http://en.wikipedia.org/wiki/Multi-level_cell

http://techreport.com/review/24841/introducing-the-ssd-endurance-experiment

http://techreport.com/review/26523/the-ssd-endurance-experiment-casualties-on-the-way-to-a-petabyte

http://www.storagesearch.com/ssdmyths-endurance.html