Nandflash lifetime calculation

Richard Weinberger richard.weinberger at gmail.com
Fri Sep 16 07:20:50 PDT 2016


Jaap,

On Fri, Sep 16, 2016 at 1:34 PM, Jaap de Jong <jaap.dejong at nedap.com> wrote:
> Hi All!
>
> I'm wondering if my expected nandflash lifetime calculation is, well at
> least, a bit accurate.
>
> So I have this nandflash and about half of it is occupied by static files.
> It will not change.
> The other half will be used by logs and databases, the dynamic part of this
> UBI filesystem.
> Their length will not change because they sort of act as circular buffers.
> Let's say this system ran a year and the maximum erase counter is now at
> 1.000.
> The nandflash is specified at 100.000 cycles and is 512MB in size.
> I added a factor of 0.7 for overhead.
>
> So I have a few assumptions, please correct me when and where I'm wrong:
>
>     1 - total space available for writing such a nandflash = 0.7 * 100.000 *
> 512MB before it reaches
>         its maximum if wear levelling is across the full device (global,
> which UBI claims)
>         --> 35840 GB
>
>     2 - when the maximum erase counter is at 1.000 in 1 year and 256MB is
> being used for that,
>         a total of about 1.000 * 256MB is written in 1 year
>         --> 256 GB
>
>     3 - the expected lifetime of this nandflash with this usage then would
> be 35840 / 256
>         --> 140 yr
>
> Would that make any sense? (Except that it is far too long to even get close
> to that....)

Well, your calculation considers only one factor, the maximum number of P/E
cycles.
Please also keep read disturb and data retention in mind.
And the environment also matters, mostly temperature.

-- 
Thanks,
//richard



More information about the linux-mtd mailing list