Why would an enterprise flash storage (SSD) have shorter data retention rate than consumer flash storage (SSD)?

[kangaroo123]

The below image comparison (from ~2010) shows that during the end of its life an enterprise SSD would have much worse data retention duration than a retail/consumer one.

This is counterintuitive to what I'd expect since enterprise SSDs are expected to be more robust.. Explanation?

 

original image:

http://www.extremetech.com/wp-content/uploads/2015/05/NAND-Data-Retention-640x392.jpg

[whispous]

One reason is for any retention needs, they'll be using slower but higher capacity hard drives instead. The SSDs are not going to be used for general data storage.

[kangaroo123]

3 minutes ago, whispous said:

One reason is for any retention needs, they'll be using slower but higher capacity hard drives instead. The SSDs are not going to be used for general data storage.

You mean that for example consumer SSD would be SLC while enterprise SSD would use QLC? I don't know if this is the case since usually enterprise SSDs have much higher P/E cycles (~100k) which requires it to have lower bit density (SLC have higher P/E cycles than QLC).

[whispous]

Just now, kangaroo123 said:

You mean that for example consumer SSD would be SLC while enterprise SSD would use QLC? I don't know if this is the case since usually enterprise SSDs have much higher P/E cycles (~100k) which requires it to have lower bit density (SLC have higher P/E cycles than QLC).

I'm not sure about which one uses which tech

[mariushm]

You're not interpreting the data correctly. 

 

A datacenter SSD won't be powered off, it will run 24/7, so the controller may actively refresh cells constantly.... How many WEEKS the data is kept while the SSD is powered off is not really relevant for a datacenter SSD.

 

A regular home user will have 25-30c ambient temperatures (air around the ssd case) when the PC is off, and the air inside the case may go up to 40-50c when in use (in a poor ventilation case).  So the chart makes sense .. 30c for power off, 40c for in use, at the intersection you get 52 weeks 

 

Note that the datacenter chart assumes higher power off temperatures and higher power on temperatures consistent with higher usage - they assume a datacenter ssd will have heavier mix of reading and writing, as ssds will be used for caching stuff, reading lots of files, they'll be used for IO... otherwise mechanical drives would be used.

 

The temperatures also depend on how the servers are cooled in the rack. Some racks have cold air in front and hot air in the back so if the SSDs are in front they'll be cooler... other racks don't have hot-cold isles, and maybe use simpler cooling systems. 

 

The datacenter SSDs may also have their firmwares optimized for less idle maintainance, less refreshing of cells, because it's assumed the SSDs will be mostly on.

If you look on the chart at power off temperature of 30c just like consumer drives, you basically go up to 50 weeks retention vs 52 weeks for consumer drives, so the datacenter SSDs actually have about the same retention.

 

 

[Tegneren]

What @mariushmsaid.

TLDR: Enterprice SSD's are made to be constantly powered on and hard use, while consumer SSD's are made to be powered off for long stretches of time and lightly used

[kangaroo123]

Thanks!

Someone has also pointed out to me that the consumer standard is for 8 hours/day activity while the enterprise standard is for 24 hours/day.