[ale] no putting swap on ssd

Ron Frazier atllinuxenthinfo at c3energy.com
Tue Feb 8 11:41:45 EST 2011


Hi Greg,

I'll admit you're a bit above my head here, but I'll tell you what I 
know.  Don't use SpinRite on an SSD in read / write mode.  Steve does 
not recommend it for this purpose.  It was designed before SSD's were 
invented for spinning platters, and it's custom tuned for their 
characteristics.  Spinrite beats up the drive, in a healthy way, by 
reading inverting writing reading inverting writing every sector of 
data, thus refreshing all the magnetic fields to maximum strength.  If 
data is hard to read, it tries exhaustively to read the sector up to 
2000 times coming at it from many various starting points, which 
minutely affect the head velocity and alignment over the track.  This 
often allows total or partial recovery of a bad sector.

Beating an SSD to death will, literally, beat it to death over time.  So 
the sector refresh process would probably be detrimental to the drive.  
I don't think we know yet if SSD's suffer from their version of bit 
rot.  However, Steve's documentation says that the magnetic fields on 
HDD's definitely do decay over time.  (He doesn't call it bit rot, 
that's a term I found on the net, but it seems to fit.)  So, the sector 
refresh process will probably not have any useful effect in terms of 
data readability.  The data recovery process, while probably not harmful 
since it's doing massive numbers of reads, not writes, would probably 
not be effective.  I would THINK that you can either read an SSD sector 
or you cannot.  I don't know for sure.  You might be able to derive some 
benefit on a bad drive where the drive is working and the data is 
scrambled due to a power failure or something by doing the SpinRite read 
only analysys (I think it's level 2 in his menu).  Since there's no head 
and no tracks, every faulty sector might read the same every one of 
those 2000 tries.  That would defeat the purpose of his statistical 
analysis.

He says he's considering doing something that relates to SSD's, but 
hasn't announced any products yet.  You may want to send him your 
suggestions at http://grc.com/feedback .  He gets so many emails he 
cannot reply personally to all, and usually doesn't.  He says he reads 
them all though, and he may respond on the podcast or simply take it 
under advisement.

Sincerely,

Ron


On 02/08/2011 10:24 AM, Greg Freemyer wrote:
> On Tue, Feb 8, 2011 at 4:23 AM, Ron Frazier
> <snip>
>    
>> I also, a few times a year, SpinRite my hard drives and refresh the data
>> in every single sector with read invert write read invert write cycles.
>> Keeps 'em purring like a kitten.  We had a long discussion about that
>> recently on the list.
>>
>>      
> Ron,
>
> Does spinrite also have a SDD refresh feature?  The issue I mentioned
> about static data being held out of the wear-leveling process would be
> overcome if all active data on the drive were read and re-written
> every so often.
>
> ie. When you write a EB the current lightly used EB should be moved to
> the free area and one of the most heavily used should be pulled and
> put into that static slot.  Then a few months later you can do it
> again.
>
> Thus a SSD refresh would be a very nice addition.
>
> Trouble is because of trim, you really only want to read/write EBs
> that are allocated.  And the ATA-8 spec. doesn't provide a way to
> tell.
>
> (eg. reading / writing a unallocated EB will turn it into a allocated
> EB.  That's fine for real data, but very bad to just do across the
> entire SSD.)
>
> Thus you have to parse the filesystem and figure out which blocks have
> live data and which don't.
>
> I think the e2fsprogs package has some userspace tools that could help
> with that, but I don't know.  (That package supports ext2/ext3/ext4.
> For xfs, use xfsprogs.  Not sure about other filesystems.)
>
> If spinrite had a tool for that, it would be really nice.
>
> Greg
>
>    

-- 

(PS - If you email me and don't get a quick response, you might want to
call on the phone.  I get about 300 emails per day from alternate energy
mailing lists and such.  I don't always see new messages very quickly.)

Ron Frazier

770-205-9422 (O)   Leave a message.
linuxdude AT c3energy.com



More information about the Ale mailing list