I've had a certain number of reliability issues with various ssd's, 
having crucial m3 sata2 ssd's and now using an adata sata3 unit.  First 
time running the m3's I ran in md raid 0, knowing a bump would kill it. 
  I found that out after about 8 months or so.  Rebuilt as a raid1 from 
then on, but had to rebuild the os (good thing for a nas).

That lasted about another 6mo or so, and one disk just stopped living. 
Good thing for raid1 now, offloaded my data, and moved to the adata units.

Those I'm still using about 2yr later, and toss the raid randomly 
probably about every 3 months.  No issues otherwise, and usually my raid 
glitch that breaks the disks is ati video driver/gl crash related.  It 
does ugly things to the raid md set, enough that I've gotten alternating 
partitions to fall out of raid oddly.  Talk about treading lightly with 
a software rebuild, but I've never *not* been able to resync the raid 
with them and run again for an indefinite period.

Performance-wise, the raid0 was smoking fast for using ubuntu, and the 
raid1 with sata2 disks wasn't noticeable worse than the raid1.  Longest 
part of boot is the bios, ata bus probing, and luks encryption - maybe 
5-8 sec to boot to login otherwise.

I did notice a big increase going from a c2d-based system with sata2 to 
a sandybridge proc/mobo with sata3 ports on it with the adata disks.  I 
watch my desktop like a hawk with gkrellm visible at all times if my 
system lags - I can generally tell what is doing it, and it's _never_ 
the disks in iowait.

Extracting or using chunks of disk with the ssd raid locally in 
real-world cases are very fast, way faster than anything on a esata 
spindle disk on the same system.

I did take great care to align the sectors as best I could find with 
ubuntu, and still not ever really sure how much it may or may not be 
accurate or "right".  I layer md, luks encryption, and lvm for fs 
flexibility, so in there I'm certain I lose trim ability, so I can't 
expect much with it.  I tuned each per recommendation for the ^2 
geometry of it, and really haven't seen half the issues as with 4kb 
spindle drives.  I'd be happy to share the recipe - works for me mostly*.

So far I'm pretty happy with these adata s511 disks even using the 
sandforce chips where others complained.  I think this was a "rev2" 
chipset if I remember right, which is probably half the battle as the 
first batch was flawed.

Anything "industrial" is just sickeningly expensive for decent consumer 
use, so I'll just buy 2.  I can buy 5-8:1 consumer for industrial ssd's 
cost for raid1+hotspare(s).

ymmv.

* My laptop died recently not able to decrypt either disk on my laptop 
using the exact same recipe as the said desktop.  Not terribly happy, 
but still a bit crosseyed to delve into it as it's unneeded at the 
moment and I'd replicated data recently.  I suspect the crappy hp 
hardware/bios, not the disks.

-mb


On 04/02/2013 12:27 PM, Derek Trotter wrote:
> I have a question about SSDs.  I've read that they like the USB thumb
> drives can be written to a certain number of times before they fail.
> What is the expected lifetime of an SSD?  They're terribly expensive if
> they're only going to last 2 or 3 years.
>
> Derek
>
---------------------------------------------------
PLUG-discuss mailing list - PLUG-discuss@lists.phxlinux.org
To subscribe, unsubscribe, or to change your mail settings:
http://lists.phxlinux.org/mailman/listinfo/plug-discuss