Question 3 vital questions about Samsung SSDs

Status
Not open for further replies.
Dec 9, 2019
13
0
10
I have been trying to find out some things about the new Samsung drives, specifically the 2TB model MZ-76E2T0B/AM.

Samsung have been extremely unhelpful, and most recently when they deign to answer they are saying that the information I want is confidential.

Clearly this is rubbish, as I am asking (see below) for stuff that is readily available via simple tests. I don't know (or care) exactly what they are trying to conceal, perhaps simply that they don't know the answer themselves. Although simple, the tests will/would take some time and expense so I am hoping that someone here has run them already, or is interested enough and flush enough to do so.

These are the questions I need answers to:


(1) how long do you expect (or specify) a powered-down device to resist data loss.

(2) how long must a device, after such a maximum power outage, be powered up to return to a fully "silvered" state.

(3) and importantly whether this re-silvering will take place within a device that is in standby state (the state where a hard disk would spin down).


Number (3) of course includes the issue of whether a device may be left in standby longer than the period mentioned in (1) above without data loss.

This information is necessary because I wish to deploy such a device in an application where it will be powered down occasionally, and where it will ideally be in standby most of the time.

If anyone can help, I appreciate it very much.

.../Leigh Clayton, Toronto Canada
 
Dec 9, 2019
13
0
10
What do you mean by "a fully "silvered" state " ?
What does that refer to?

Thanks for the response.

Please pardon the length of this response, I am not good at brevity at the best of times and it wasn't clear from your post how much detail you needed.

I take the term from discussions about HDD-based file systems (and I believe that they take it from things like astronomical primary mirrors, which occasionally need to be re-coated with a "silver" coating - actually aluminum I think - to keep them in ideal reflecting condition).

What I mean by "fully silvered" is that the data on the drive is all freshly written, so that any ongoing decay has been eliminated.

As far as I can tell Samsung doesn't document this, but it is of course well-known that flash memory is not really nonvolatile. It is actually "slowly volatile". SDHC cards keep their data for a long time - years, certainly, and the figure I see for most similar memory is 15-20 years. SSD s tend to be much more volatile, due to the compactness of the physical bit-cells. The MLC SSDs, such as the ones I'm asking about, do not (as far as I can tell) physically retain data for anything like as long as SDHC flash. The figures I've seen range from a year to 6 months, although as I've said Samsung seems to have circled their wagons about this detail and I've never seen or heard anything from them concerning this issue. Thus my questions.

The fundamental problem is that the bits are stored as voltage ranges within semiconductor cells/wells, and the electrons that implement these voltages are not absolutely constrained to remain in their places; they slowly dissipate as time passes.

I am a long-time user of Samsung SSDs, and I have never lost a single bit on any of them. Since the physics of these products ordains that they do lose data continually at the atomic/die level, I am forced to conclude that they employ some sort of active refresh logic. Obviously they would not want to refresh needlessly, since that would increase the product's susceptibility to failure due to write endurance limits. I am happy to leave the details of whatever compromise they have worked out to them; they clearly have it just about right for drives left powered up all the time, as my existing drives tend to be.

However, the application I am looking at would ideally leave the drives in standby mode most of the time, and their specs note that these SSDs draw a lot less power in standby mode. This made me wonder if the active resilvering logic, that I infer must be there, is disabled in this mode. Plus, as I mentioned, these devices will be powered down at times. What I am trying to find out is exactly what parameters I will need to enforce on these drives to avoid data loss.

Obviously I can simply buy a few drives and test them, but this will mean an outlay for the drives and at least several months, possibly years, to run the tests. These are both things I would like to avoid. I confess I am exceedingly irked at Samsung for avoiding my questions (at this point they are not even responding to my messages), but that is irrelevant to my question and I have tried to keep that out of these posts. ( Of course, if these tests take long enough, Optane may become available first, in a form able to solve my problem. Intel has confirmed that Xpoint/Optane memory does not have this issue in any form - as one might expect from considering its very different physical implementation.)

I hope that this is clear, if not exactly concise.

.../Leigh
 

USAFRet

Titan
Moderator
I've not seen any reports on 'power off' endurance.
There was one a few years ago that scared quite a few people, in that "SSD's can start to lose data after a week of power off"
Then you read, and...
The drive needed to be past its actual TBW limit (not the warranty number)
It needed to be 'full'
It needed to be kept in an abnormally hot environment.

What is the time period you're looking at for this unpowered state?
 
Dec 9, 2019
13
0
10
I've not seen any reports on 'power off' endurance.
There was one a few years ago that scared quite a few people, in that "SSD's can start to lose data after a week of power off"
Then you read, and...
The drive needed to be past its actual TBW limit (not the warranty number)
It needed to be 'full'
It needed to be kept in an abnormally hot environment.

What is the time period you're looking at for this unpowered state?

Thanks for responding. I am aware of the factors to which you refer, all of them contribute to the general frailness of flash storage, and thus my concern to deploy this particular device properly, to minimize my likelihood of encountering such problems.

(I am very concerned, actually, about how much of our infrastructure resides in flash disks, or more importantly, in the firmware storage of devices monitoring the health of things like nuclear power stations. )

I am not looking for any specific time period.

My goal is purely to discover how to use one of these devices in a way that keeps it operating perfectly - ie no data loss.

As I said, the application I am looking at right now (a data server of sorts with currently about 10 TB of duplicated hard disk dataspace) is one where the present disks are kept either in standby (spun down) or actually powered off almost all the time. I am considering moving part of the data to SSD because that will permit faster access after a period of inactivity, and importantly to me, it will remove the increased likelihood of failure associated with unusually high HDD activity. (I am aware that this last point is unproven and contentious, but it weighs on my mind when I do use one of them an unusual amount.)

Perhaps I am being naive, but I don't think it is so unreasonable to want to understand this aspect of these devices, and I really don't think Samsung is offering proper support to me to deny me this information. But I don't see any way to dig it out of them at this point. The company is clearly in trouble of some sort, presumably over some of the problems they've had with their phones, and this has made them uncomfortable saying anything publicly, perhaps without their International CEO signing off on it? (I am, as I admitted, rather peeved at them and I'm starting to think it would be best to stay away from any of their products. Irrational perhaps, but if they are this unreasonably uncommunicative about SSDs, I have to think they are being just as reticent about their other products.)

Pardon me for sharing the annoyance, it's been eating at me.

Of the three questions, by far the most important to me is whether the period after which data loss becomes likely / not unlikely is affected by keeping the device in standby. As I think I said, their specs show considerably less power usage in this state, so it makes me wonder exactly how much of the internal housekeeping we're talking about is shut down in this mode. I could of course eliminate standby mode for whichever of these devices I deploy, but without knowing if there is any point to this I'm unhappy to (possibly) waste all that power.

..../Leigh
 
Dec 9, 2019
13
0
10
It's only 10TB of data. Why not use a SSD based storage for quick access and a stringent back-up routine to HDD?

Yes, yes .. there are many ways I could go from where I am, all have their advantages and disadvantages. That is not my issue. As a matter of fact, I am not concerned about actually losing data forever, as I DO back up everything that needs backing up. My data is all automatically duplicated across a set of volumes - the 10TB figure is the amount of data-space, not the amount of storage across the arrays.

In any case, I am not asking for ways to work around this issue. I assure you that I am perfectly capable of coming up with any number such, including leaving things exactly as they are. I am hoping, rather, that I can somehow learn what Samsung won't tell me, so I can create the solution that suits me. It is not, after all, that I am asking for anything complicated or unreasonable, at least as I see it.

If I am going to shell out money for one (or perhaps several) of these things, does it really seem reasonable to you that I should simply and blindly hope that I choose the right way to deploy them? The point of moving my data to SSD is so that I'll be able to use it more conveniently and more confidently. If my confidence comes from keeping my fingers crossed, and knowing that I can rebuild my data whenever it goes missing, I haven't really gained enough to make the expense and trouble worthwhile.

.../Leigh
 

USAFRet

Titan
Moderator
Bottom line - I've not read any actual real tests of "write once, leave on the shelf for X years, and is the data gone or is it still there."

Possibly because we've not yet reached any number of "years" that these things have been in existence.
Similarly, we don't see that same thing for spinning drives. Or DVD (except for the theoretical archive disks, which are supposed to last several decades).

That number does not exist, because the drives in question have not existed for that long. Any test you might see would be forced aging, which is not real.
 
Yeah, I think something like this would be difficult to test, and the results would likely depend on storage conditions. And it's probably something that would take multiple years of remaining in a powered-down state, by which point the drive might no longer be on the market, and newer drives with newer memory chips with different characteristics will have taken their place. It's also likely not just a simple number, but rather an increasing probability of a bit flipping over time. And the exact tolerances might even vary from one batch of chips to the next.

Does Samsung have at least a rough idea of what the rate of degradation will be? Probably, but they also probably can't provide a hard number stating that a given drive will assuredly retain data in a powered down state for a given number of months or years.
 
Dec 9, 2019
13
0
10
Bottom line - I've not read any actual real tests of "write once, leave on the shelf for X years, and is the data gone or is it still there."

Possibly because we've not yet reached any number of "years" that these things have been in existence.
Similarly, we don't see that same thing for spinning drives. Or DVD (except for the theoretical archive disks, which are supposed to last several decades).

That number does not exist, because the drives in question have not existed for that long. Any test you might see would be forced aging, which is not real.

Thanks for your reply; see reply to this and next post below.
 
Dec 9, 2019
13
0
10
Bottom line - I've not read any actual real tests of "write once, leave on the shelf for X years, and is the data gone or is it still there."


Possibly because we've not yet reached any number of "years" that these things have been in existence.

Similarly, we don't see that same thing for spinning drives. Or DVD (except for the theoretical archive disks, which are supposed to last several decades).


That number does not exist, because the drives in question have not existed for that long. Any test you might see would be forced aging, which is not real.





Yeah, I think something like this would be difficult to test, and the results would likely depend on storage conditions. And it's probably something that would take multiple years of remaining in a powered-down state, by which point the drive might no longer be on the market, and newer drives with newer memory chips with different characteristics will have taken their place. It's also likely not just a simple number, but rather an increasing probability of a bit flipping over time. And the exact tolerances might even vary from one batch of chips to the next.

Does Samsung have at least a rough idea of what the rate of degradation will be? Probably, but they also probably can't provide a hard number stating that a given drive will assuredly retain data in a powered down state for a given number of months or years.

Thanks for both of your replies.

I assure you both that this has occurred to me, but if what you say is the case, why has Samsung refused to say so? Not, of course, that they don't know how long data will stay valid in a powered-off drive, although they must have internal data on that. (It would be a function or graph, of course, as you say it is not a single number).

But I have been careful to ask about a drive with only a single write (that is, full capacity written exactly once), so I was not asking for the full graph.

Also, as I have said both to them and here, the most important thing for me is the difference between a drive powered down and in standby. If your implication is correct, that is that there is no active re-silvering happening in any of these drives, they could easily have just said so. This would mean that standby and powered-on behaviour (and powered-off behaviour) would be identical. That would have been very easy to state in some form,
and presumably they would have no reason to conceal that fact.

I really dislike being in the position of accusing them of deceit or covering-up anything, but I don't see any other explanation other than simple malice.

Probably Goethe said:
Do not attribute to malice anything that can be satisfactorily explained by incompetence.

.../Leigh
 

USAFRet

Titan
Moderator
An SSD, even in a powered on state, does not rewrite, or 'resilver', the data unless the system causes some write or delete action to that drive.
Powered off, it simply remains as is. Eventually, I imagine some cells would lose voltage. I have no idea what that powered off timespan would be.


But there are two concepts with SSD's that would not apply with HDD's and this 're-silvering'.

Wear leveling
It distributes write cycles across all cells in the drive, to ensure individual cells do not expire early.

TRIM
https://www.digitalcitizen.life/simple-questions-what-trim-ssds-why-it-useful
 
Dec 9, 2019
13
0
10
An SSD, even in a powered on state, does not rewrite, or 'resilver', the data unless the system causes some write or delete action to that drive.
Powered off, it simply remains as is. Eventually, I imagine some cells would lose voltage. I have no idea what that powered off timespan would be.


But there are two concepts with SSD's that would not apply with HDD's and this 're-silvering'.

Wear leveling
It distributes write cycles across all cells in the drive, to ensure individual cells do not expire early.

TRIM
https://www.digitalcitizen.life/simple-questions-what-trim-ssds-why-it-useful

I am well aware of both the concepts you mention. As I have said, I am a long-time user of flash SSD, especially Samsung devices, and I only use Linux/FreeBSD so I am forced to do a little more configuration manually than some MS users may be. Neither of the things you mention apply to a drive which is filled up once and never subsequently modified, which is a very good approximation to the application I am considering.

Your post sounds as if it is meant to be authoritative, but may I ask whence the authority? I mean no offense, but what you have said leads to the inescapable conclusion that data on an SSD will last exactly as long if the device is powered off and placed in a safety deposit box, as if it is kept powered on in a computer. This is contrary to everything I've heard, although I have no personal direct contrary evidence since I have never tried it. (Flash is of course not like CMOS, for example, where a constant power source will keep data safe - there is so far as I am aware no such passive option conceivable for flash.)

I have to wonder, though. If this is really the case, why does Samsung refuse to say it to me? Clearly if they HAD said that, I would have gone away happy and never bothered you all here.
(Actually, one of their first-level people DID say that to me, sort of. They said that the data would stay good FOREVER, which would be nice. But I know a bit too much physics to be able to believe that, and none of the others have ever said anything like it.)

.../Leigh
 

USAFRet

Titan
Moderator
No, it is not meant to be authoritative. Just providing information that may not have been known.
Nor do I personally know the actual lifetime of data on an unpowered SSD.

As I said:
"Eventually, I imagine some cells would lose voltage. I have no idea what that powered off timespan would be. "

However, I doubt there is some nefarious cabal within Samsung or the solid state memory industry keeping that data from you.
It may simply because they do not know.
No one can state that a device will retain 100% of its data for 20 years, if that device has only existed for 5 years.
 
There is no such thing as a drive that will be able to keep your data forever unless you're talking about the tiny glass disk that can store up to 360 terabytes of information and will be able to survive for billions of years without damage or data loss.

And like USAFRet said it's hard to know what will happen in 20 years if the drive only been existing for 5.
 

4745454b

Titan
Moderator
Just wondering out loud here, does any of this even matter to you? From what I've read in your posts, these drives will be powered off for maybe a few months at a time. Standby should "re silver" as you put it then continue their normal low level of power draw. Unless you are planning on powering down the SSD for years at a time I simply don't see why you'd care. It will get powered back up after X months time frame so why be bothered by this?

I'm sure Samsung has a reason for not telling you/us. Trade secrets or something like that. I don't see deceit, malice, incompetence , etc. Sometimes we do have to wonder about things.
 
Just out of curiosity, I pulled an old 250GB OCZ SSD out of my parts drawer. That drive has not been powered in MORE than two years. Likely, it's been more like three or four. Connected it, all data is there and is fine. So, I seriously doubt the premise that these drives lose data any faster than any other flash based devices.

Even the VERY BEST optical discs don't claim to be able to keep data uncorrupted for more than a few years although we KNOW that they do last longer, but a fair measure. Nobody is going to tell you it is guaranteed though, especially not the disc manufacturers.

It think this information, in general, is pretty accurate and is a good "yardstick" if you don't want to get caught with your pants down.


CD and DVD

According to the US National Archives, CDs and DVDs have very similar lifespans. Generally, unrecorded (blank) CDs and DVDs have a shelf-life of five to ten years. The experiential life expectancy of recorded CDs and DVDs is between two and five years, though based on manufacturer claims, ten to twenty five years, or even longer, isn’t unprecedented. In any case, using very conservative numbers will reduce the risk of losing data.


These numbers also depend on environmental factors and how often you use the disc. Any optical media is extremely susceptible to damage because there is little protection on the readable surface—just think about how many CDs of yours have been scratched through regular use, it happens to all of us.

Blu-Ray

Verbatim writeable Blu-ray disks come with a lifetime warranty, though I couldn’t find any reliable info on how long they supposedly retain data (or how long other brands of Blu-ray discs last). Under prime environmental conditions, they supposedly last quite a bit longer than CDs and DVDs because the method for recording data results in more durable storage, but even though they likely last quite a bit longer, they’re still optical media, which means they’re susceptible to scratching, high temperatures, and sunlight, just as the others.

M-Disc

The M-Disc is an optical media storage disc that is a supposedly “permanent storage solution”. There are claims that it may be able to last up to 1,000 years, even in the face of environmental damage caused by scratching and high temperatures. While the M-Disc is a new format, it can be used with any standard DVD drive to read information, but since the data is engraved into advanced metals, a special M-Disc-ready drive is required to write it. Plus, since this technology is so new, the 1,000 year lifespan is only theoretical so only time will tell how long these advanced discs will really last, though they do have some fairly impressive research backing their theories.

Hard disk drives

Most hard disk drives (HDD) last between three and five years before some component fails. That doesn’t always mean the drive is irrecoverably busted. But three to five years is still about how long they last, whether you’re talking about an internal drive for a server or desktop, or an external hard disk drive. With all of the moving parts inside, something will eventually stop working. But as with any media storing important data, it’s important to use quality hardware.

Flash storage

Flash storage comes in three different common storage media: Flash drives, SD cards, and solid state drives (SSDs). eHow says flash drives can last up to ten years, but as mentioned on NYTimes.com, flash memory doesn’t usually degrade because of its age, but rather because of the number of write cycles, which means the more you delete and write new information, the more quickly the memory in the device will start to degrade. Since all these devices are similar in that they all use flash memory, they’ll all degrade in a similar fashion. However, one thing is certain: better hardware will pay off. Given the variety of manufacturers, lifespan might differ quite a bit from one device to another, but flash memory devices rated for more write cycles will usually last longer. When it comes to flash drives and SD cards, you’ll likely lose them or ruin them in the washing machine before anything else happens.
 
Last edited:

rubix_1011

Contributing Writer
Moderator
I have an older 120GB SSD that I put into a USB enclosure to use as portable storage. It hasn't been connected for at least 2 years (mostly because I have USB sticks of 64GB and larger and no longer need it).

All data was there and able to be read fine.

Are we certain that this data degradation/missing/unreadable isn't due to something else? It would seem this thread is dead set on proving this theory when it could be something else. I would even question the drives being stored next to something with a magnetic charge over data loss due to normal atomic attrition.
 
Dec 9, 2019
13
0
10
No, it is not meant to be authoritative. Just providing information that may not have been known.
Nor do I personally know the actual lifetime of data on an unpowered SSD.

As I said:
"Eventually, I imagine some cells would lose voltage. I have no idea what that powered off timespan would be. "

However, I doubt there is some nefarious cabal within Samsung or the solid state memory industry keeping that data from you.
It may simply because they do not know.
No one can state that a device will retain 100% of its data for 20 years, if that device has only existed for 5 years.

Thanks got your reply.

I assure you I would accuse Samsung of no such thing if they had responded in any of the ways that they could have, if what you have said is true.

If they had said "we do not actively re-silver data on our SSDs", or "data written to our SSDs will persist for as long in standby or completely powered off as it will if the device is kept powered up" or even "we do not know for certain how long the data will persist but leaving the device in standby will not affect this" then I would have gone away happy. It was them telling me (twice) that they would not respond to my questions for internal confidentiality reasons, and subsequently ignoring me, that led me to wonder what was going on. Amd led me to hope someone here could help me.

.../Leigh
 
Dec 9, 2019
13
0
10
There is no such thing as a drive that will be able to keep your data forever unless you're talking about the tiny glass disk that can store up to 360 terabytes of information and will be able to survive for billions of years without damage or data loss.

And like USAFRet said it's hard to know what will happen in 20 years if the drive only been existing for 5.

Yes but I have not asked them how long the drive will keep my data, I have been asking them about the difference I will see depending upon whether I keep the drive powered up, or in standby.
 

USAFRet

Titan
Moderator
"they would not respond to my questions for internal confidentiality reasons "
Then that is your answer.
Not everything is publicly announced. Corporate secrets.

If you asked Intel/Crucial/Sandisk the same thing, you'd probably get the same sort of answer.
 
Dec 9, 2019
13
0
10
Just wondering out loud here, does any of this even matter to you? From what I've read in your posts, these drives will be powered off for maybe a few months at a time. Standby should "re silver" as you put it then continue their normal low level of power draw. Unless you are planning on powering down the SSD for years at a time I simply don't see why you'd care. It will get powered back up after X months time frame so why be bothered by this?

I'm sure Samsung has a reason for not telling you/us. Trade secrets or something like that. I don't see deceit, malice, incompetence , etc. Sometimes we do have to wonder about things.

Thanks for the response.

Your response makes several assumptions - eg that there is a resilvering process, and that it will run OK in standby mode.

I have been looking at a new, 2TB, MLC drive and it is reasonable, I think, to wonder what, if anything, is sacrificed to get the capacity. The specs they do publish show a marked decrease in power consumption in standby mode, which is what led me to ask the question originally. I am sure someone within Samsung has a reason for keeping this secret, and if they'd answered my questions I would have been able to make an informed decision and that would have been that. As I've remarked here, those questions are ALL about things which anyone who owned such a drive could find out for themselves if they invested the time and money to run the simple tests some of you are describing.

As for "why be bothered", I began this by just trying to discover the best way to use the darned thing. I had no intention, I assure you, of getting this involved or pestering either Samsung or you people.

...../Leigh
 
Dec 9, 2019
13
0
10
Just out of curiosity, I pulled an old 250GB OCZ SSD out of my parts drawer. That drive has not been powered in MORE than two years. Likely, it's been more like three or four. Connected it, all data is there and is fine. So, I seriously doubt the premise that these drives lose data any faster than any other flash based devices.

Even the VERY BEST optical discs don't claim to be able to keep data uncorrupted for more than a few years although we KNOW that they do last longer, but a fair measure. Nobody is going to tell you it is guaranteed though, especially not the disc manufacturers.

It think this information, in general, is pretty accurate and is a good "yardstick" if you don't want to get caught with your pants down.

Thanks for your post. I am quite familiar with the things you say about CDs, DVDs, hard disks, tape, etc. I have worked in the mainframe industry for almost 50 years and managing data across various media has been part of what we did.

As I have said previously here, I have NOT asked them, or you, for an estimate of how long my data will last on their devices. I have run several Samsung SSDs over the last few years, all are still in service, and I have so far as I know never lost a bit from any of them. I have been asking about the difference between the behaviour of one specific device in powered-up mode as opposed to standby.

.../Leigh
 
Dec 9, 2019
13
0
10
I have an older 120GB SSD that I put into a USB enclosure to use as portable storage. It hasn't been connected for at least 2 years (mostly because I have USB sticks of 64GB and larger and no longer need it).

All data was there and able to be read fine.

Are we certain that this data degradation/missing/unreadable isn't due to something else? It would seem this thread is dead set on proving this theory when it could be something else. I would even question the drives being stored next to something with a magnetic charge over data loss due to normal atomic attrition.

Thanks for your post, and for running the test.

As I have tried to make clear, I have NOT lost any data personally, and I did not ask them or you about data loss, specifically, or ask for any estimates as to how long my data will last if the device is used correctly.

What I asked them, and what I hoped someone here could tell me, is simply how best to deploy these new, much more compact, devices.

.../Leigh
 

USAFRet

Titan
Moderator
What I asked them, and what I hoped someone here could tell me, is simply how best to deploy these new, much more compact, devices.

.../Leigh
"How to deploy" depends on what you're going to be doing with it.

Home PC, corporate desktop, database server, movie production, backup drives, archival storage....all would specify different methods and different specific drives.
 
Status
Not open for further replies.