Testing Power Loss Scenarios on SSDs: Does any review site actually do this?

[generalmx]

https://www.usenix.org/system/files/conference/fast13/fast13-final80.pdf has some awesome info about how many SSDs, even Enterprise-level SSDs (MLC or TLC), don't handle power-loss scenarios as well as enterprise-level HDDs; even with the much-touted super/ultra-capacitors (that can provide a little extra time to prevent data corruption). While the chance is tiny, I think benchmarking & review sites should make the effort. I'm not even sure, say, the Crucial M500's inclusion of supercapacitors, is unequivocally a good thing (did most of you even know the M500 is one of the very few consumer drives to include them?): the super/ultracaps could be used to make the standard write caching all drives do much more aggressive (since you have more wiggle-room on power-loss), thus improving performance.

The study linked is quite detailed on how they tested the SSDs, and I would love to see some benchmarking and/or review site derive a test from their methods. SSDs (even cheaper ones) are so damn fast now, we need testing on features like this to help us really differentiate both Enterprise and Consumer.

Here's a decent news article citing the study: http://www.infoworld.com/t/solid-st...isk-massive-data-loss-researchers-warn-213715

And here's a slightly more independent source, Wikipedia, mentioning problems like lower page corruption: https://en.wikipedia.org/wiki/Solid-state_drive#Battery_or_super_capacitor

 

[Cerb]

The problem is that it's hard to test well, since some given drive might just be better or worse with that testing method or platform.

 

[generalmx]

The problem is that it's hard to test well, since some given drive might just be better or worse with that testing method or platform.

That's certainly true; however, the same could be said about any testing or even benchmarking procedure (which is why the smart people look at a range of benchmarks and not just, say, how some hardware does with the latest Battlefield engine). Anyway, the paper I linked outlines an empirically-driven test with multiple citations to backup why they tested the way they did, and how it's significant: you can't get much better than that. Of course, the paper may be biased, but either way, it's ultimately up to the reader to make the distinction of significance to them.

 

[William Gaatjes]

That is something i would also like to see.
But indeed hard to test. That means a switch to interrupt the power to the SSD.
I do not know for sure but that could damage the sata link. With E-sata it should be possible, i think.

 

[Puffnstuff]

Another reason that I don't think that a pc is complete unless it has a battery backup with avr.

 

[Zaxx]

This is a little OT but thought I'd mention the Crucial's M500 enthusiast SSD's have caps for power protection. This is the only consumer grade ssd I know of to offer power protection. I'm currently waiting for a M500 1TB (960GB) that I won from ssd review to arrive. I may try yanking the power connector during a file transfer but not too sure how to objectively test. Hmmm...

 

[Hellhammer]

The problem is that it's hard to test well, since some given drive might just be better or worse with that testing method or platform.

Exactly. If we had a reliable way to test this, of course we would do this. However, at least I've not figured out any convenient method. The method used in the paper would require special hardware and likely some EE background as well to be done properly.

 

[William Gaatjes]

Perhaps this device is capable of power failure tests to test the SSD :

What types of devices can it test? SATA 3/6Gbps, SAS 3/6/12Gbps, FC 1/2/4/8/16Gbps, STP 3/6Gbps, and PCIe based protocols like NVMe and SCSI3 for starters but there are probably more that I missed. The OakGate tester can do protocol conformance testing, stress testing, error correction testing, I/O load testing, retention after power cycles, and even check the validity of arbitrarily amounts of recently written data. Or it can do combinations of the above, in short about everything you would need. It can also pound a drive to failure automatically while logging every write.

http://semiaccurate.com/2013/12/23/oakgate-makes-ssd-tester/

 

[William Gaatjes]

This is a little OT but thought I'd mention the Crucial's M500 enthusiast SSD's have caps for power protection. This is the only consumer grade ssd I know of to offer power protection. I'm currently waiting for a M500 1TB (960GB) that I won from ssd review to arrive. I may try yanking the power connector during a file transfer but not too sure how to objectively test. Hmmm...

Would that not damage the sata interface of the MB ? Be careful...
To be safe, i think you should try it on a hot pluggable sata interface. Should be able to withstand a power loss event by removing the connector.

 

[TripleThreatTLT]

The problem is you're looking for a needle in a haystack with this testing, at least on a drive with the backup caps. To test it properly you need 50+ drives and to asynchronously pull the power on the drives 100's of times. The failure rate is very small. I don't think its feasible for most review sites to do.

I will also say specialty hardware is needed. Although the simple way to is just to buy a network controlled power strip and cut power to the entire pc.