lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael McCandless <luc...@mikemccandless.com>
Subject Re: Lucene tests killed one other SSD - Policeman Jenkins
Date Mon, 19 Aug 2013 17:22:35 GMT
I see, so the 70X is how soon you can expect to see your first
reallocated sector?

But, that still seems way too low; the underlying erase/write cycles
on the NAND cells is in the 1000s I think?  Hmm looks like for modern
20nm NAND cells it's ~1000.

If the wear leveling is working why should you see reallocated sectors
after only 70X writes?  My OCZ Vertex 3 is at 341X and no reallocated
sectors so far!

Mike McCandless

http://blog.mikemccandless.com


On Mon, Aug 19, 2013 at 1:09 PM,  <karl.wright@here.com> wrote:
> Right, that's what I said.  And one write means writing the *whole* disk.  So Mike and
I may *both* be right. ;-)
>
> Karl
>
> -----Original Message-----
> From: ext Uwe Schindler [mailto:uwe@thetaphi.de]
> Sent: Monday, August 19, 2013 1:07 PM
> To: dev@lucene.apache.org
> Subject: RE: Lucene tests killed one other SSD - Policeman Jenkins
>
> Hi,
>
> I misunderstood this number on the first time, too.
>
> 70x does not mean 70 times write fully and then the SSD is gone. It means 70 times write
of the SSD capacity until the first error somewhere in the flash memory occurs. Those errors
are not fatal, they are just list in the "relocated sector count" or decrease this  "SSD remaining
lifetime" smart attribute. So 70 might be correct.
>
> Uwe
>
> -----
> Uwe Schindler
> H.-H.-Meier-Allee 63, D-28213 Bremen
> http://www.thetaphi.de
> eMail: uwe@thetaphi.de
>
>
>> -----Original Message-----
>> From: Michael McCandless [mailto:lucene@mikemccandless.com]
>> Sent: Monday, August 19, 2013 6:57 PM
>> To: Lucene/Solr dev
>> Subject: Re: Lucene tests killed one other SSD - Policeman Jenkins
>>
>> I am curious :)  70X is really way too low.
>>
>> I just checked the OCZ Vertex 3 behind the nightly Lucene performance
>> tests ( http://people.apache.org/~mikemccand/lucenebench/index.html ),
>> which indexes the full Wikpedia English export 3 times nightly.
>>
>> It's written 74.4 TB, read 35.7 TB, and its odometer (SSD_Life_Left
>> from
>> smartctl) says 98 out of 100 ... which I think equates to ~11,000x -
>> 17,000x write endurance.
>>
>> This thread is also fun:
>>
>>     http://www.xtremesystems.org/forums/showthread.php?271063-SSD-
>> Write-Endurance-25nm-Vs-34nm
>>
>> Users destroy their SSDs by writing and writing and writing until it
>> dies :)
>>
>> Mike McCandless
>>
>> http://blog.mikemccandless.com
>>
>>
>> On Mon, Aug 19, 2013 at 12:50 PM,  <karl.wright@here.com> wrote:
>> > Mike, I'm talking about a 1TB SSD option for some hardware we are buying.
>> If you are really curious, I can ask the people who are doing the
>> project for the model and specs.
>> >
>> > Karl
>> >
>> > -----Original Message-----
>> > From: ext Michael McCandless [mailto:lucene@mikemccandless.com]
>> > Sent: Monday, August 19, 2013 12:47 PM
>> > To: Lucene/Solr dev
>> > Subject: Re: Lucene tests killed one other SSD - Policeman Jenkins
>> >
>> > Er, which drives have write endurance only 70x?  That's exceptionally low?
>> I thought this was more like ~1000x for modern consumer SSDs?
>> >
>> > Mike McCandless
>> >
>> > http://blog.mikemccandless.com
>> >
>> >
>> > On Mon, Aug 19, 2013 at 12:35 PM,  <karl.wright@here.com> wrote:
>> >> " Only 70 full writes seems a little bit low for an SSD."
>> >>
>> >> That's what I thought.  I was astounded to learn that that is in
>> >> fact correct
>> (at least for some of the drives we are using here).  Automatic
>> recovery is how the SSD copes with this failure rate.
>> >>
>> >> But it is entirely possible that the cause of your problems are
>> >> something
>> else.
>> >>
>> >> Karl
>> >>
>> >> -----Original Message-----
>> >> From: ext Uwe Schindler [mailto:uwe@thetaphi.de]
>> >> Sent: Monday, August 19, 2013 12:31 PM
>> >> To: dev@lucene.apache.org
>> >> Subject: RE: Lucene tests killed one other SSD - Policeman Jenkins
>> >>
>> >> Hi Karl,
>> >>
>> >> Only 70 full writes seems a little bit low for an SSD. This SSD has
>> >> a capacity
>> of 64 Gigabytes so 53 Terabytes written is not bad for it. It already
>> had the first few errors, so you might be right, as you see from this Smart data:
>> >>
>> >> 231 SSD_Life_Left           0x0013   096   096   010    Pre-fail  Always
      -       0
>> >>
>> >> So this "arbitrary" counter just went down from 100 to 096. I have
>> >> the
>> feeling that the current issues with this SSD are not really write
>> problems, more hardware interface or firmware problems (as seen quite
>> often for OCZ Vertex 2). Otherwise it would fail very early. According
>> this data it should survive more 10th of years?
>> >>
>> >> Uwe
>> >>
>> >> -----
>> >> Uwe Schindler
>> >> H.-H.-Meier-Allee 63, D-28213 Bremen http://www.thetaphi.de
>> >> eMail: uwe@thetaphi.de
>> >>
>> >>
>> >>> -----Original Message-----
>> >>> From: karl.wright@here.com [mailto:karl.wright@here.com]
>> >>> Sent: Monday, August 19, 2013 5:07 PM
>> >>> To: dev@lucene.apache.org
>> >>> Subject: RE: Lucene tests killed one other SSD - Policeman Jenkins
>> >>>
>> >>> I am told that SSD's are spec'd for only 70 full writes before they
>> >>> get an error.  The error block is set aside but eventually something
>> >>> critical gets hit.  So you should probably should expect this to happen
>> again.
>> >>>
>> >>> Karl
>> >>
>> >>
>> >>
>> >> ---------------------------------------------------------------------
>> >> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org For
>> >> additional commands, e-mail: dev-help@lucene.apache.org
>> >>
>> >
>> > ---------------------------------------------------------------------
>> > To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org For
>> > additional commands, e-mail: dev-help@lucene.apache.org
>> >
>> >
>> > ---------------------------------------------------------------------
>> > To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org For
>> > additional commands, e-mail: dev-help@lucene.apache.org
>> >
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org For additional
>> commands, e-mail: dev-help@lucene.apache.org
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
> For additional commands, e-mail: dev-help@lucene.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message