[Frieder Ferlemann <frieder.ferlemann@we...>]

Hi again,

Am 04.03.2010 18:38, schrieb Frieder Ferlemann:
> Thanks!
>
> 198 Read_Sectors_Tot_Ct     0x0000   ---   ---   ---    Old_age   Offline      -       63412716
> 199 Write_Sectors_Tot_Ct    0x0000   ---   ---   ---    Old_age   Offline      -       93996899
> 200 Read_Commands_Tot_Ct    0x0000   ---   ---   ---    Old_age   Offline      -       1352556
> 201 Write_Commands_Tot_Ct   0x0000   ---   ---   ---    Old_age   Offline      -       1953112
>
> I'm surprised about this one:
>
> 208 Average_Erase_Count     0x0000   ---   ---   ---    Old_age   Offline      -       50
>
> because to my reading (on the wrong assumption of only linear writes)
> this would translate to 50 * 64GByte drive capacity = 3 TByte
> having been written.
> (the ratio of about 60 between 3 TByte and 93996899 sectors *
> 512 bytes/sector = 48 GByte would seem very high to me)
>
> (write sectors being higher than read sectors could be plausible
> because I did two installs of linux. First one was just for
> fun a milestone version of openSUSE 11.3 which I a few hours
> later exchanged for 11.2)
>
> I bought the drive as a new device and I did no firmware update
> as it already had the latest firmware 1916.
> The first say 10 hours the disk was used with a 2.6.31 kernel
> since then a kernel of 2.6.33 series was installed.
>
> I intend to do a follow up posting if the drive has some more
> power on hours.



actually I did never expect to report back this early,
but I'm quite upset about the Average Erase Count
of 72 of the drive.

198 Read_Sectors_Tot_Ct     0x0000   ---   ---   ---    Old_age   Offline      -       82208608
199 Write_Sectors_Tot_Ct    0x0000   ---   ---   ---    Old_age   Offline      -       106676155
200 Read_Commands_Tot_Ct    0x0000   ---   ---   ---    Old_age   Offline      -       1781513
201 Write_Commands_Tot_Ct   0x0000   ---   ---   ---    Old_age   Offline      -       2275882

208 Average_Erase_Count     0x0000   ---   ---   ---    Old_age   Offline      -       72

Meanwhile 6492 MByte have been written    (106676155 - 93996899) Sectors * 512 Bytes /Sector
while  1408511 MByte have been erased!    (72 - 50) * 64.023.257.088 Bytes

This would mean that wear leaveling (or garbage collection/whatever)
causes an overhead of more than 200 in flash erase cycles!!!

Intolerable.

I hope there is a mistake in my calculations or somewhere else.


My configuration:
- The system has been in what I'd call "light desktop use"
- The SSD was shipped with the latest firmware 1916
- I did never mount swap on it (4GB Ram and Swap on
  a conventional disk, so no need to)
- The root partition was mounted noatime
- Using a kernel of 2.6.33 series

More detail:
/dev/sda2 on /     type ext4 (rw,noatime,acl,user_xattr)
/dev/sda1 on /home type ext4 (rw,acl,user_xattr)
Linux version 2.6.33-29-desktop (geeko@...) (gcc version 4.4.1 [gcc-4_4-branch revision 150839] (SUSE Linux) ) #1 SMP PREEMPT 2010-03-05 10:49:02 +0100
Output of smartctl -a /dev/sda is attached


If my observations can be independently confirmed this
should probably escalate (LKML / LWN / Slashdot)

Can someone else have a look?

Greetings and thanks,
Frieder