I have read that zeroing the SSD helps with restoring speed. If those claims 
have any validity I am not sure. I think perhaps if you never ran the disc 
using the discard mount flag, or  if you formatted this disc with an older 
version of ext4, it might actually help to zero the disk.

>From my experience using xfs/ext4 discs without the discard flag/trim enabled 
the performance will degrade over time (only tested with older SSDs, over a 
year ago).  I believe xfs enables FITRIM (non realtime trim) automagically if 
the device supports trim, however I am unsure about ext4.

Check result of to see if trim is working at the block level.
cat /sys/block/sda/queue/discard_max_bytes

For reference here are my 2 SSD in my desktop.

SATA3 - 309.89 MB/sec
SATA2 - 186.91 MB/sec

On Tue, 04 Feb 2014 14:26:32 C. Falconer wrote:

> Derek Smithies wrote, On 04/02/14 13:02:
> >  Disk read/write times are (my view) somewhat misleading.
> >
> >Big old snip<
> 
> Sorry - don't really care about the how or why, was just after a simple
> rough and ready metric to see if my SSD was excessively slow.
> 
> Steve's numbers suggest that 180 MBytes/second is indeed sub-par, and
> similar measurements I got from the guys in #nzlinux on the freenode IRC
> channel confirm that my SSD is slow.
> 
> Now I have to see if it was always that slow... My recollection is that
> it did get 270 Mbytes/second when it was new, so I will investigate if
> trim is working as expected.   Could also be the running load so I'll go
> single user and do some more testing
> 
> > On 04/02/14 12:31, Steve Holdoway wrote:
> >>   Timing buffered disk reads: 970 MB in  3.01 seconds = 322.75 MB/sec
> >>   Timing buffered disk reads: 870 MB in  3.00 seconds = 289.73 MB/sec
_______________________________________________
Linux-users mailing list
[email protected]
http://lists.canterbury.ac.nz/mailman/listinfo/linux-users

Reply via email to