Dealing with SSD's inevitable crashes?

[stephens241]

Like many users, I am interested in using SSDs due to speed and heat issues. However, it is my understanding that unlike hard-drives, it is only a matter of when, rather than if, the SSD will fail to cull data as required. Also, the failure problem increases at higher memory capacities. So, I'm curious to know how ThinkPad users here are planning to manage this problem --frequent backups to a portable hard-drive, for example, or some other solution? Has anyone had this problem yet and successfully addressed it, and if so, any tips? Is it possible to have an external hard drive, with the operating system on board, by which to boot the machine in the event of a SSD meltdown? Sorry if these seem like stupid questions, but I'm definitely a neophyte with this stuff.

[mgo]

...it is only a matter of when, rather than if, the SSD will fail to cull data as required. Also, the failure problem increases at higher memory capacities. So, I'm curious to know how ThinkPad users here are planning to manage this problem .

Could you site the source of this information about SSD failures?

Since any drive can fail, I just always have an image available with a emergency boot CD or boot USB drive.

Also, why not just keep your old mechanical drive around, and if the SSD goes bad, simply slip in the replacement.

I am using four solid state drives and so far, nothing has gone wrong... But then, I am a bit of an "experimenter" and have prepared for any little "emergencies".

[visionviper]

Like many users, I am interested in using SSDs due to speed and heat issues. However, it is my understanding that unlike hard-drives, it is only a matter of when, rather than if, the SSD will fail to cull data as required. Also, the failure problem increases at higher memory capacities. So, I'm curious to know how ThinkPad users here are planning to manage this problem --frequent backups to a portable hard-drive, for example, or some other solution? Has anyone had this problem yet and successfully addressed it, and if so, any tips? Is it possible to have an external hard drive, with the operating system on board, by which to boot the machine in the event of a SSD meltdown? Sorry if these seem like stupid questions, but I'm definitely a neophyte with this stuff.

Yes, SSDs are in a way a ticking time bomb in the way that they are pretty much guaranteed to fail - but this failure is so many years down the line you will either:
A) Have a new laptop
B) Have a new SSD because the new one has more space/speed or whatever.

The estimates I have seen put a SSD's lifetime upwards of a decade or more. How many hard drives do you still use from 10 years ago?

[stephens241]

Well, the estimates to time of failure range from 5 to 10 years. So far as I can tell, these estimates are not accounting for memory size, whereas the time to failure is directly correlated with size.

What's getting lost in all the excitement about SSD, somehow, is that the failure is not a matter of 'if' but 'when'. Also, an estimated range to failure spanning 5 to 10 years is not especially enlightening or reassuring. Will one's own SSD fail in 5 years, 10 years or maybe 3 years? Obviously, the number of rewrites contributes to the time of failure. But so does exposure to lapsed power sources and other factors.

Just to be clear, I'm not trying to rain on anyone's SSD parade. I just want to be intelligently prepared to deal with the risks.

So, is it possible with the T400 ThinkPad line to back up, boot up, and most importantly, run in complete duplicate, with an external hard-drive (i.e. with a backup copy of one's operating system), should the failure arrive when I'm on, say, an airplane?

[ssd_thinkpad]

I am using ssds for years without having ever any problems.

There is only one problem regarding ssds: If there would be any hardware failure, you can not recover the data (for reasonable costs).

[qviri]

.

[ssd_thinkpad]

I would have imagined that, somewhat akin to magnetic drives, you slowly lose capacity as you wear out individual memory chips or cells and the firmware reassigns or removes them from use. Is the binary fail issue connected with the controller PCB, then? I'd have thought that as reliable as any other PCB in a computer under the same circumstances.

That maybe true for lots of mlc drives. I am not sure if it is for slc drives. My slc drive is never slow - either when full or half empty, I do not notice any speed difference.

[stephens241]

"I am not very educated in this issue. Could someone explain to me how SSDs fail?"

There are 3 issues. 1) SSD withstands a limited number of rewrites and then stops being "rewriteable". 2) The rewrites are more subject to corruption if initiated during power lapses. 3) As the drives sustains more and more rewrites, the speed of the drive decreases.

[qviri]

.

[LondonConsultant]

Right, but none of these sound like they'd cause a sudden death without warning? Do SSDs report SMART information on their health like magnetics do?

If you know it's coming, you can prepare better.

Note that over a third of hard drive failures occur without any warning, ie no SMART error signals at all. The reference is an excellent Google paper: http://labs.google.com/papers/disk_failures.pdf

Also, perhaps this topic on SSD crashes should be in the General Hardware/Software forum, rather than the T400/T500 forum.

[AvalonXIII]

The death of SSD is defined as "no longer rewriteable" aka you can still read the data from the drive, and copy it over to a HDD before throwing it away. The average MLC SSD is expected to last 10 years at a rate of 10GB of write per day, while SLC SSD is expected to last 10 times that amount, or 100 years at a rate of 10GB of write per day. So, as you can see, it's on par, if not better than some HDD out there. Obviously, not all SSD will last as long as that, but you can say the same thing about HDDs.
So the death of SSD is exaggerated. I've been using SSDs for two years now, and they're all going strong.
And no, the bigger the size of the SSD, the longer it will last, not the opposite. The job of the controller in the SSD is to distribute writes equally to all cells, such that when you delete data from a cell, it actually marks the cell's data no longer usable, but doesn't delete the data. So when a new write happens, the controller will write the data to a completely new cell. When all the new cell runs out, then it starts to write back to the "marked as deleted" cells. As you can see, the more cells you have, the longer your SSDs will last. (There are other concerns such as performance degradation due to actually deleting a used cell before rewriting it, but that's for another topic)

[deforest]

Well as a new owner of a ssd drive, that also crashed within a week of owning it, I have a hp whs (windows home server), that I use to back up all my computers.

I booted my machine with a livecd and found that I still had a partition, but when I ran chkdsk, it can back with it being empty, i.e. the partition was full of free space with no directories or files.

So if I didn't have a back up, my data would have been toast.

[AvalonXIII]

Well as a new owner of a ssd drive, that also crashed within a week of owning it, I have a hp whs (windows home server), that I use to back up all my computers.

I booted my machine with a livecd and found that I still had a partition, but when I ran chkdsk, it can back with it being empty, i.e. the partition was full of free space with no directories or files.

So if I didn't have a back up, my data would have been toast.

Sounds like a faulty controller. In some instance, HDDerase helps bring the controller back to life. Have you tried the software yet?

[rkawakami]

I am not very educated in this issue. Could someone explain to me how SSDs fail?

The basic problem is that current solid state drives (SSD) are built with Flash memory technology. Flash, by nature, can only withstand so many erase/write cycles for each cell (well, block of memory really), after which that block will no longer be able to correctly store some data. Typical endurance for Flash memory is between 100,000 and 1,000,000 erase/write cycles; that's the number of times in which any single memory block can be erased (written to all ones) and then re-programmed. The controller inside the SSD is designed to evenly spread out the memory blocks which are being used - "wear leveling" is the term usually used to describe this process.

I never much liked the idea of using Flash memory as a disk drive in the early days (and still don't for the relatively slow write performance aspect). However, with improved chip processes and intelligent memory management, the current crop of SSDs are probably just as reliable as magnetic disk platters (if not more since there's no physical movement involved).

[bill bolton]

the current crop of SSDs are probably just as reliable as magnetic disk platters

Depending on the particular application the computing platform is being used for, probably may be too great a risk, and that is the issue at hand that stephens241 is trying to get some rational discussion going on.

I personally suspect that mass market SSD technology has not yet reached a point of adequate maturity where there is any good response possible..... except to watch what happens to the generally perceived risk profile over time.

Those who have experience, in real-time, with the long journey of rotational disk technology from the initial "flying head" washing-machine sized drives, through to the point of having a reliable mass market "black box" consumer level product, will understand what that is about

Cheers,

Bill B.

[rkawakami]

The estimates I have seen put a SSD's lifetime upwards of a decade or more. How many hard drives do you still use from 10 years ago?

Well, let's see... 1999 or before.... My 600Mhz PIII Gateway desktop which shipped in Nov of '99 still has the original 20GB Quantum Fireball. Also in that system (but only 9 years old) is a Maxtor 45GB drive which is still operational. Granted, I don't use that system all that much any more but the drives are still working. However, if you are talking about laptop drives then I think that I've tossed out several minuscule (<6GB) DARA and DJSA units from some of my 600 series systems over the years. There may be a couple that I have that are around 8 years old but I can't be certain unless I start pulling drives out. But given that a laptop drive sees more physical shocks than a desktop system, a relatively short lifespan is to be expected from that. It's the one (and so far, maybe "only") advantage that an SSD has over rotational disks.

The company that I'm currently contracting for (Grandis, Inc.) is working on a type of magnetic memory called STT-RAM. That's right, magnetic, but not the old "core" memory from 40 years ago. This is memory cells about the same size as DRAM and smaller than Flash. I don't fully understand all of the physics involved but basically it comes down to the ability to effect a change in some magnetic material in two different orientations. That is the basis for the storage of zeros and ones. The best part of this is that like Flash, it's a non-volatile memory. It does not need the continuous application of power in order to retain data. And unlike Flash, it does not appear to wear out, you can re-write a single bit without having to erase an entire block of memory and it writes as fast as it reads (in the order of a few nanoseconds). An interesting technology which may be ready for "primetime" in a couple of years.

[twister6]

As an alternative to keep from writing too often to your SSD, use it as a program drive while a "spinning" hdd in ultra-bay for storage and data/document files That's what I did with my T400s, 128GB SSD for program and Scorpio black in a caddy for storage.

[returnzer0]

I was almost upset when I first started reading this thread because there just seemed to be so much FUD going around (yeah yeah I know, arguing on the internet). Being skeptical of new technology is a good thing, but spreading misinformation is something else entirely.

First, the flash inside of SSDs are basically EEPROMs, Electronically Erasable Programmable Read Only Memory, which are extremely tolerant of reads but not so much of the erase-then-write cycle. What this means is that when an SSD fails, the vast majority of the time it will simply turn into a read only flash drive and you will be able to get your data off of it.

The erase/write cycles are what limits the lifetime of an SSD because every time a cell is erased, it wears it out a little. A flash memory block can be written to about 10,000 times on the low end for MLC and 100,000 on the high end for SLC (in the lab, SLC has been made with 1 million cycles but no products have come out based on it yet). Most SSDs are sold with extra flash in them so that when the blocks do fail, the data can be reallocated accordingly without a loss in accessible capacity. The memory controllers also have wear leveling algorithms that keep track of how many times a block has been written to and ensures they get worn down uniformly. With the combination of spare flash and wear leveling, the write cycles become a non-issue for all but the most write-intensive applications.

Intel says that their first generation 80GB X25-M, their consumer MLC drive, will not fail before five years with more than 100GB/day of erase and write cycles. Think about that for a second. You would have to completely fill and then erase the drive every day for five years straight to possibly reach the point where the drive turns into a read-only brick. With Intel's 32GB X25-E, their enterprise class SLC drive, you'd have to write 2.22TB every day for 5 years to achieve possible failure. Even then, years after all the blocks have worn out, you can still read the data off the drive as long as the controller is still good. Also keep in mind that, as mentioned previously, this will only increase as you move up to higher capacities and the technology itself is improved.

I'm sure I don't have to explain why SSDs just make sense in laptops, especially when you're targeting the thin and light market. No moving parts, less weight, more easily adaptable to smaller form factors, etc. Apart from that, I don't know how anyone can deny the advantages of sub millisecond access times, faster sequential transfer rates, ridiculous IOPS, and random access transfer rates that were unheard of in the consumer market. As far as I'm concerned, the only advantage that traditional HDDs have over SSDs is price/GB, but who's to say that won't change in the future too.

Specifically to the OP, it doesn't matter what kind technology you are using to store your data; it is always a matter of "when" the drive will fail. It's not as if you know exactly when your magnetic HDD is going to fail and can prepare accordingly. I would almost venture to say that you're in a better off position with an SSD because you actually can get a rough estimate of when it will fail for you based upon your usage. Generally though, I don't see how you could treat data backups any differently while using SSDs than what you should already be doing with traditional spinning HDDs.

I highly encourage anyone who's at even remotely interested in SSDs to read up on Anand's very informative article. It's long and really delves into how the technology works, but it makes an excellent primer for anyone considering an SSD.
http://www.anandtech.com/storage/showdo ... i=3531&p=1

[mgo]

Being skeptical of new technology is a good thing, but spreading misinformation is something else entirely.

Generally though, I don't see how you could treat data backups any differently while using SSDs than what you should already be doing with traditional spinning HDDs.

Thanks for the helpful remarks about SSDs. Since I am using one in a T60 ThinkPad, I'm curious about your thoughts on:

-disable superfetch
-disable prefetch
-disable defragmenter
-no pagefile

Other posters in other fora have suggested some strange sounding tweaks that do not seem to even relate to a solid state drive. Some Microsoft people seem to claim that if Windows 7 sees a SSD, it will make changes to match, but that is not my experience.

thanks...

[AvalonXIII]

I almost forgot another possiblity of failure, which is controller failure. Yes, this has happened to many Intel drives, where the capacity is read as 8MB and you can't write to or read from the drive.
However, controller is less likely to fail than your typical flash chip so it'll be fine.

[returnzer0]

-disable superfetch
-disable prefetch
-disable defragmenter
-no pagefile

Other posters in other fora have suggested some strange sounding tweaks that do not seem to even relate to a solid state drive. Some Microsoft people seem to claim that if Windows 7 sees a SSD, it will make changes to match, but that is not my experience.

A lot of those options are meant to help compensate for the shortcomings of spinning HDDs, and since those same shortcoming don't apply to SSDs, they do not need to be enabled while using one. I'm not familiar with the differences between superfetch and prefetch, but they are there to help minimize the effect of the slow random access speeds of HDDs. Since SSDs have such low access times, applications are much quicker to load even without prefetching, though I see no real problem with leaving it enabled as it shouldn't hinder performance.

Defragmenting a HDD made it faster by rearranging data on the physical platters so that it more closely resembled the filesystem. It attempts to make it so that chunks of a file are located sequentially, all the files in a directory are adjacent to one another, etc, and once again the purpose of this is to try to minimize the effects of the random access times. SSDs can access any chunk of any file, no matter where it is on the disk, relatively instantaneously, and thus defragging has no effect on performance. The only thing that degramenting an SSD will do is wear it out quicker, so it should be disabled.

The pagefile has more to do with your RAM than your storage. If you have sufficient RAM, then you never want to be using a pagefile because it will always be slower. If you need to use a pagefile, running it off an SSD will be much quicker because it mimics RAM much better than traditional HDDs. Heavy use of a pagefile on an SSD will wear it out faster. I have no information on this, but it would be interesting to know the volume of data that gets written to memory in a day, since the only scenario I can think of where limited erase cycles might become a problem for the average consumer is if it were set up as a giant pagefile. Personally, I disable my pagefiles. I have at least 4GB of RAM in the computers I use which seems to be adequate for most tasks. However, I have run into cases when a program will not run because it detects no pagefile, and I know that Microsoft's OSs cannot write out a memory dump without a pagefile.

I don't know in which version it was implemented, or even if it's been implemented yet, but the story is that when Win7 detects an SSD, it disables scheduled defragmenting and enables the TRIM command. I'm not aware of any other SSD specific changes that it makes.

[mgo]

[quote="returnzer0The pagefile has more to do with your RAM than your storage.[/quote]

I think I'll go with 200 meg -upper and lower- of swapfile (to keep Windows from yelling at me) and see how it works, along with disabling superfetch and prefetch and defrag. Thanks for the reply...

[stephens241]

"I was almost upset when I first started reading this thread because there just seemed to be so much FUD going around (yeah yeah I know, arguing on the internet). Being skeptical of new technology is a good thing, but spreading misinformation is something else entirely."

Revisiting this thread late as I went out of town as it developed. Sorry you were "almost upset", but I find the comment quoted above wildly out of line with the tone and intent of my initial inquiry. Precisely what "misinformation" am I "spreading" in my initial post?

When solid state drives fail, the failure probably entails complete inaccessibility. That is not necessarily the case with failures that plague hard-drives. Actually, I've yet to ever experience a failure in which all data was forevermore inaccessible with a hard-drive, on any of the computers I've purchased. I'm sure complete inaccessibility happens with hard-drives, but I know from experience that the inevitability factor is not the same as what's been documented with SSDs. To equate the potential problems of a SSD with a HD is simply an inaccurate rendering of the obvious distinctions between these technologies.

But all of this is beside the point of my initial inquiry. My question concerned the logistical issues/problems of backing-up and booting-up once an SSD has been fried. With the embedded security features of the ThinkPad, are there any special problems with backing one's system on a conventional portable hard-drive and booting-up with the hard-drive? I haven't ever dealt with this scenario and I'm wondering what problems, if any, to expect, from users who may already have experienced this kind of problem.

You haven't experienced such a problem? Great! That wasn't my question.

[mgo]

My question concerned the logistical issues/problems of backing-up and booting-up once an SSD has been fried. With the embedded security features of the ThinkPad, are there any special problems with backing one's system on a conventional portable hard-drive and booting-up with the hard-drive? I haven't ever dealt with this scenario and I'm wondering what problems, if any, to expect, from users who may already have experienced this kind of problem.

My tactic with a solid state drive is the same as with a rotating hard drive; I create an image using one of the commercial imaging programs.

I run my operating system and programs on a 30 gig partition on the ssd, and my data is on the second ssd partition. I image the system partition, and use Robocopy to back up the data on the second ssd partition.

The only difference is that the backups and imaging take about half the time using a solid state drive, compared to a rotating drive, even when the destination drive is a rotator.

Should the SSD become inaccessable, I can use a bootable thumb drive or bootable CD to lay down the image C:\ to another drive, and Robocopy wil restore the other data to the new drive's 2nd partition.

[returnzer0]

Revisiting this thread late as I went out of town as it developed. Sorry you were "almost upset", but I find the comment quoted above wildly out of line with the tone and intent of my initial inquiry. Precisely what "misinformation" am I "spreading" in my initial post?

I wasn't singling you out. There's a lot of misinformation in this thread and I just wanted to clear it up. However, since you insist...

When solid state drives fail, the failure probably entails complete inaccessibility.

Wrong. If you actually read through my post, you'd know why.

That is not necessarily the case with failures that plague hard-drives. Actually, I've yet to ever experience a failure in which all data was forevermore inaccessible with a hard-drive, on any of the computers I've purchased.

You haven't experienced such a problem? Great! I just hope you don't think that it's a rare occurrence. You should read up about the bathtub curve and how it relates to warranties.

I'm sure complete inaccessibility happens with hard-drives, but I know from experience that the inevitability factor is not the same as what's been documented with SSDs.

The inevitability factor? of failure? It is 100% for all drives. You made this mistake in your original post as well. ALL drives will fail.

To equate the potential problems of a SSD with a HD is simply an inaccurate rendering of the obvious distinctions between these technologies.

You are inaccurately extrapolating the results of your single anecdote to the whole SSD market. Just because you had one bad experience doesn't mean that all SSDs are faulty. You seem to apply this concept correctly to HDDs in realizing that your experiences are uncommon, but all logic disappears once you move on to SSDs.

But all of this is beside the point of my initial inquiry. My question concerned the logistical issues/problems of backing-up and booting-up once an SSD has been fried.

I did specifically answer this. Going from an HDD to an SSD shouldn't change your backup habits.

With the embedded security features of the ThinkPad, are there any special problems with backing one's system on a conventional portable hard-drive and booting-up with the hard-drive?

No. There are no problems unique to SSDs in this regard.

[returnzer0]

Hey mgo,

You were asking about some Windows specific tweaks that could be made when using SSDs, and there was a pretty good article published recently on that topic here:
http://www.bit-tech.net/hardware/storag ... or-vista/1

It seems like you've got your bases pretty well covered though.

[mgo]

Hey mgo,

You were asking about some Windows specific tweaks that could be made when using SSDs, and there was a pretty good article published recently on that topic here:
http://www.bit-tech.net/hardware/storag ... or-vista/1

It seems like you've got your bases pretty well covered though.

Many thanks for the link.
-mg