Studies of consumer SSD integrity after power failures?

[Jambe]

Hi all! I'm wondering if any of you know of robust analysis of how consumer SSDs respond to power failures.

I'm also trying to determine the value of the cap banks in Intel's 320 series and Crucial's M500 series, and whether the caps are appreciable purchase considerations for users of desktops without UPSes.

 

[fffblackmage]

I voted yes, although really, it's not the most important feature I look for in a consumer SSD. I imagine ratings and price is most important for most people. If data integrity is that important, you'd most likely be looking at enterprise SSDs instead anyways.

 

[Mark R]

There has been a study into data integrity on consumer SSDs, which show that the vast majority will corrupt data under crash/power failure conditions.

Even some SSDs with super-capacitor backup will corrupt data or brick, in the event of power failure.

 

[MrPickins]

I like the idea of super-caps in my drive, but question whether SSDs are any more prone to data corruption during a power failure than HDDs.

In either case, if I can't afford data corruption, I use a UPS.

 

[Drako]

Absolutely needed, unless you don't care about data corruption.

 

[Burner27]

Edit: I didn't see 'without UPSes' in original post. My bad.

 

[sxr7171]

Yeah i think it's important. I would choose an SSD with than without.

 

[Socio]

SCSI and SAS controller manufactures have small built in or addon batteries to prevent such data loss and corruption.

If motherboard manufactures and SATA controller manufactures would implement something like that for onboard SATA or SATA controller cards it would be a fairly inexpensive solution and you would not need a UPS.

 

[Mark R]

Yeah i think it's important. I would choose an SSD with than without.

Fair point, if the drives with capacitors actually work as specified.

Studies gave shown that at consumer level, capacitor equipped drives are no better than non-equipped drives at avoiding corruption /bricking.

 

[Jambe]

Studies gave shown that at consumer level, capacitor equipped drives are no better than non-equipped drives at avoiding corruption /bricking.

Could you link these "studies"?

I've only been able to find this one:

Understanding the Robustness of SSDs under Power Fault

Alas, given that only four of the fifteen SSDs tested were known to feature power loss protection, and given that the study only mentions "supercapacitors", I'm guessing all the tested loss-protection drives were enterprise-class (as opposed to the consumer 320 and M500 lines which use banks of regular capacitors, not one or more supercapacitors).

 

[Red Squirrel]

I would spend money on buying a UPS before spending extra money on something to prevent a problem that can be prevented with a UPS. Another problem it will prevent is if you are in the middle of working on a file or playing a game you can at least finish the round before shutting down.

I never realized it was an issue with SSDs though. Do they actually brick, or does the last bit of data that was being written simply end up corrupted?

UPSes are actually not that expensive, it amazes me a lot of people don't have one. You don't need to go all out either, a <$100 cheapy one will still be better than none at all.

 

[Mark R]

I never realized it was an issue with SSDs though. Do they actually brick, or does the last bit of data that was being written simply end up corrupted?

The study linked above explains it.

Very simply, they found 5 ways in which the SSDs could behave following a power failure:
1. All data complete and correct up to the point of power failure (optimal behavior)
2. Most recently written data missing or incomplete (Acceptable behavior)
3. Last written data OK, but earlier writes missing or corrupted (bad behavior as this will defeat the data protection mechanisms of journaling filesystems, and database logfiles which rely on the drive perfoming writes in a guaranteed order)
4. Massive random data corruption.
5. Drive bricked.

Could you link these "studies"?

MY mistake. Had 2 copies of that 1 study titled differently in my pile of notes.

 

[Drako]

LOL at the term "shorn writes" from that sudy.

 

[Cerb]

SCSI and SAS controller manufactures have small built in or addon batteries to prevent such data loss and corruption.

If motherboard manufactures and SATA controller manufactures would implement something like that for onboard SATA or SATA controller cards it would be a fairly inexpensive solution and you would not need a UPS.

It really wouldn't be that cheap. Consumers simply do not service their computers, so capacitors are really the only viable option.

I never realized it was an issue with SSDs though. Do they actually brick, or does the last bit of data that was being written simply end up corrupted?

The two are not exclusive. What LBA it is writing to could very well be part of the data it is writing.

 

[fzabkar]

Alas, given that only four of the fifteen SSDs tested were known to feature power loss protection, and given that the study only mentions "supercapacitors", I'm guessing all the tested loss-protection drives were enterprise-class (as opposed to the consumer 320 and M500 lines which use banks of regular capacitors, not one or more supercapacitors).

I posted a question, and analysis, in the following thread:
http://forums.storagereview.com/ind...kup-capacitors-how-much-time-before-shutdown/

Strangely, it appears that Intel's latest enterprise class SSDs use cheap 47uF aluminium electrolytic capacitors for backup, whereas their consumer grade SSDs use expensive Kemet Organic 470uF capacitors.

 

[fzabkar]

Just to put the economics into perspective, 47uF, 6.3V KO capacitors cost about US$8 - $9 each in single quantities, depending on whether one opts for the 10 milliohm or 4 milliohm ESR devices. OTOH single quantities of the KZH 47uF, 35V, low ESR aluminium electrolytic capacitors cost 35 cents each.

This means that the backup capacitor bank for the enterprise SSD would cost me $0.70, but the backup capacitors for the consumer grade SSD would set me back around $50.

http://www.digikey.com/scripts/dksearch/dksus.dll?vendor=399&keywords=T530X477M006AHE

http://au.mouser.com/United-Chemi-C...H/_/N-75hqw?P=1z0zl7sZ1yzn2jjZ1z0wrk9Z1z0x3xj

 

[Red Squirrel]

Interesting stuff... guess this stresses the importance of UPSes, and in the case of critical systems, redundant power supplies. While not as catastrophic, this type of corruption can happen with spindle drives too if shut down badly.