Jump to content


Photo

which Nand Flash SSDs are vulnerable to sudden power loss?


  • You cannot start a new topic
  • Please log in to reply
6 replies to this topic

#1 MRFS

MRFS

    Member

  • Member
  • 190 posts

Posted 04 January 2014 - 12:02 PM

I'm occasionally encountering messages posted to User Forums

that warn about total data loss if and when popular MLC and TLC SSDs

experience a sudden power loss.

 

In some instances, those SSDs actually "bricked"

(turned into non-functional "bricks").

 

For myself, I find it difficult to believe that end Users would

build a powerful workstation and then NOT power it with a

quality battery backup / UPS (uninterruptible power supply).

 

All of our workstations are powered by an APC UPS,

and that may explain why those workstations generally

do NOT experience sudden power loss;  and, believe me,

our UPS units do switch over to batteries on a regular basis

(probably because of faulty 110V wiring in our building).

 

 

I'd like to open a broader discussion here concerning the

perils of sudden power loss and the likely effects on

the current installed base of SLC, MLC and TLC SSDs.

 

 

If SSDs are failing because of sudden power loss,

the solution may be as simple as installing a

quality battery backup / UPS unit.

 

In our experience with at least a dozen workstations,

powering all PCs with a quality UPS has the predictable

effect of prolonging the useful life of the entire PC -- all components --

particularly if the UPS is paired with a quality system PSU

and the 3 primary voltages stay well within spec.

 

 

 

 

 

 


#2 MRFS

MRFS

    Member

  • Member
  • 190 posts

Posted 04 January 2014 - 04:25 PM

http://www.extremete...lling-your-ssds

 

[begin quote]

The impact of sudden power loss

The implications of this research are significant. It suggests that SSDs, including enterprise SSDs, should not be trusted to behave in the proper fashion, or to be as robust as HDDs. Indeed, the number of hits for the phrase “disappearing SSD” is huge, and while many refer to the Crucial M4, that drive is not the only one listed. I myself have run into this problem in the past, with several drives unexpectedly dying after random power cycles. I never thought to check for a wider issue until now.

 

Without vendor information, there is little practical advice to be offered.

 

The best thing a user can do is attempt to ensure that the power doesn’t unexpectedly turn off, via even a small uninterruptible power supply (UPS). Laptop users obviously have less to fear on this front, as your systems have batteries built in.

 

Even a 5-10 minute battery would be sufficient to give the user time to shut down. Manufacturers are unlikely to start talking up these issues honestly — no one wants to admit that previous products have been anything less than ironclad.

 

[end quote]

 

 

A more scientific article is here:

 

http://www.eetimes.c...?doc_id=1279443

 

Understanding the effects of power failure on flash-based SSDs
 
[begin quote]
 
As flash-based solid-state drives (SSDs) get more popular in all kinds of computing devices, the integrity of flash memory when power failure occurs becomes increasingly important. Power failure for SSD is potentially much more dangerous than it is for conventional hard drives. Because SSDs use complex flash translation layers (FTLs) to manage the mapping between logical block addresses and physical flash memory locations, if power failure corrupts the metadata about this mapping, the entire SSD can become inoperable. To ensure reliability, system designers must understand what kinds of corruption power failure can cause to design products that can withstand power failures and the resulting data corruption.
 
[end quote]
 
 
The rest of that article would benefit a lot from
proper grammatical editing, because it contains
several grammatical errors e.g. I would add a
little punctuation to this sentence:
 
"To ensure reliability, system designers must understand
what kinds of data corruption power failures can cause,
in order to design products that can withstand power failures
and prevent the resulting data corruption."
 

#3 Brian

Brian

    SR Admin

  • Admin
  • 5,388 posts

Posted 04 January 2014 - 05:08 PM

Even in an SSD with no caps on it (which is essentially all client SSDs), it's of course abnormal behavior for an SSD to brick and lose data. We've seen plenty of SSDs fail, but unexpected power loss should only kill in flight data, and that's even assuming a lot of things are going on at once and the worst of the worst outcomes occur. When we test client drives we routinely pull the power plug (unexpected power loss) from them and haven't lost a drive because of that...at leaf that we're aware of. 


Brian

Publisher- StorageReview.com
Twitter - @StorageReview

 

#4 MRFS

MRFS

    Member

  • Member
  • 190 posts

Posted 04 January 2014 - 11:46 PM

I honestly can't imagine doing professional computing

for many hours of any given work day, and NOT

powering every workstation with a quality UPS.

 

Not only should this minimize SSDs failures

due to sudden power loss;  it just makes good

common sense to protect all components

in a powerful workstation, particularly when

the UPS supplies uninterrupted AC power

to a quality system PSU, with "soft shutdown"

enabled in the UPS monitoring software.

 

Gosh, in a few large tower chassis we've

examined closely, there is often room

for 2 x ATX PSUs;  the second PSU bay

could host an AT-style PSU with its own

ON/OFF switch.

 

The latter AT-style PSU could keep

all SSDs powered ON even when the

system shuts down and all ATX SATA

power cables switch OFF, like this

prototype I built several years ago:

 

 

IMG_1569.annotated.JPG

IMG_1572.annotated.JPGIMG_1569.JPG


Edited by MRFS, 04 January 2014 - 11:56 PM.

#5 MRFS

MRFS

    Member

  • Member
  • 190 posts

Posted 05 January 2014 - 12:14 AM

Also, an AC/DC power adapter with 4-pin Molex connector

could supply ALWAYS ON power to an SSD subsystem e.g.:

 

http://www.newegg.co...N82E16812123309

 

Then a simple Molex-to-SATA power adapter like this one

could supply input power to 2 x 2.5" SSDs:

 

http://www.newegg.co...N82E16812119010

 

This Startech model has latching-style SATA power connectors:

 

http://www.newegg.co...N82E16812400307

 

 

It would be a good idea to plug the AC/DC power adapter into a power strip

with a separate ON/OFF switch, if it does not have its own ON/OFF switch.

 

As such, this adapter would plug into the power strip,

and the power strip would plug into the UPS, so that

this subsystem could be switched ON and OFF manually

and independently of the rest of the system.


Edited by MRFS, 05 January 2014 - 12:20 AM.

#6 jtsn

jtsn

    Member

  • Member
  • 94 posts

Posted 09 January 2014 - 05:17 AM

Every regular PC shutdown is more or less a "sudden power loss" for a SSD/HDD, because the PSU cuts the power to all drives without notice in advance. The OS makes sure it flushed all buffers, but then the SATA drive is on its own when dealing with the power-down.

If a drive isn't able to handle this, it is broken by design.

#7 MRFS

MRFS

    Member

  • Member
  • 190 posts

Posted 09 January 2014 - 11:31 AM

In this context, I'm quite sure that the meaning of "sudden power loss"

does NOT include a normal SHUTDOWN sequence, like Start | Turn Off Computer

with Windows. 

 

Yes, the OS makes sure it has flushed all buffers, and

that step necessarily implies that there are not any

files "in transit" when the power switches OFF.

 

At least one article I reviewed clearly stated that

I/O "in transit" is very vulnerable to corruption

insofar as it is not allowed to complete normally.

 

 

Here's a very simple analogy, with which I am

familiar from having written system software:

 

To keep this example simple, let's say there

is a very large data file, and one auxiliary file

which is an "index" into the large data file.

 

And, let's say that the large data file is

being updated with a WRITE, and that

WRITE completes but the power switches

OFF before the index file can be written too.

 

Or, the reverse sequence may occur

(depending on how it's programmed):

the index is written, but the power switches

OFF before the large data file is updated.

 

This type of corruption would be even more

complicated if there were multiple indexes,

and not all such indexes were updated properly.

 

 

Assuming that the OS properly flushes all file

system buffers before the power switches OFF,

that assumption necessarily describes something

OTHER THAN a "sudden power loss" as the

latter term is being used in the articles I've read.

 

 

Think of "sudden" as implying "accidental" or "no warning whatsoever."

 

 

Read in this article above:

 

http://www.eetimes.c...doc_id=1279443

 

 

Because SSDs use complex flash translation layers (FTLs) to manage the mapping between logical block addresses and physical flash memory locations, if power failure corrupts the metadata about this mapping, the entire SSD can become inoperable.


Edited by MRFS, 09 January 2014 - 11:35 AM.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users