BreakPoint

cancel
Showing results for 
Search instead for 
Did you mean: 

Solid State Drive Life Expectancy: White Paper Feedback

Highlighted

Hi,

I'm new here, was a lurker before.

VERY useful and supportive Forum: thanks.

 

NI has posted a very well written White Paper called: "Understanding Life Expectancy of Flash Storage) (July 23, 2020).

Understanding-life-expectancy-of-flash-storage 

 

Based on this, my questions are:

(1) References: Can the references for this be added please?

I have access to most academic literature, and would only need standard academic style references.

I'm sure that whoever wrote this can supply them.

It doesn't have to be done explicitly in the text, just a list at the end would do.

 

Although I think in future, that following the standard style of a technical paper with references given when needed would be preferable to add credibility to a document, as well as to enable it to be revised as time went on, and more was known. 

 

(2) Update: Related to this issue, this White Paper is a bit dated now, since although it's dated 20-July-2020, it does not cover 3D TLC or QLC, and the "Typical Layers of Calls" are quoted as "Varies, 64 typical", whereas since 2019 and onwards, for most SSD's this is extended to 96 layers, and will soon go higher.

 

And it does not distinguish between the two types of process used for 3D structures: Charge Trap Flash ("CTF") (most 3D NAND flash) and Floating Gate Cell ("FGC") (older 2D NAND, as well as Micron for 3D).

 

Based on my reading, I suspect that 3D TLC behaves similar to 2D MLC

 

I'm not yet sure of the differences between CTF and FGC for the same NAND type, as in how Micron 3D 64 layer TLC behaves with respect to Toshiba 3D 64 Layer TLC.

 

And I'm also not completely sure how 3D QLC behaves: I'll post separately my current estimate.

 

A major problem is NAND flash manufacturer transparency: the NAND Flash makers do not release much technical information or their newer flash, anything newer than 2D (2005) is buried under NDA's.

 

And please bear in mind I'm not looking at the very simplistic ratings of SSD drives, as given by DWPD, or TBW. Those are too broad and vague to be of much use. I like the approch of the White Paper where it used a model to compare individual Flash Cell Endurence to the overall DRIVE Endurence ("Endurence" is a deliberate mispell, since the robot seems to block the correct term??).

 

I think that if you follow the logical extension of this, then the extremely high endurence ratings based on huge DWPD for most SSD's, that lead most to assume their SSD's will last longer than a Volkswagen Beetle, are misleading and in fact they need to consider much lower values, that a significant number of users might well see, and experience data loss and/or drive failure.

 

I have some related comments/feedback, but will add in a 2nd post to keep this short.

Thanks for any replies.

 

Alan Jarvis

0 Kudos
Message 1 of 3
(270 Views)
Highlighted

Minor point, but the site robot kept on objecting to the term "Endur_ance". It said that term was "not allowed in this community". Eh?

 

I had to change it to "Endurence" to sneak it past.

0 Kudos
Message 2 of 3
(258 Views)
Highlighted

@WarthogARJ wrote:

Minor point, but the site robot kept on objecting to the term "Endur_ance". It said that term was "not allowed in this community". Eh?

 

I had to change it to "Endurence" to sneak it past.


Most likely due to a SPAM attack and blocking that word was the simplest and fastest mitigation.


GCentral
There are only two ways to tell somebody thanks: Kudos and Marked Solutions
Unofficial Forum Rules and Guidelines
0 Kudos
Message 3 of 3
(197 Views)