High-end SLC SSDs No More Reliable than MLC SSDs: Google Study

[btarunr]

A FAST '16 paper titled "Flash Reliability in Production: The Expected and the Unexpected," by Professor Bianca Schroeder of the University of Toronto, and Raghav Lagisetty and Arif Merchant of Google, studied the reliability data from millions of SSD drive-days over a period of 6 years, to come up with some very interesting conclusions on SSD reliability. One of the study's biggest findings is that high-end (read: enterprise) SSDs with single-level cell (SLC) NAND flash memory are no more reliable than cheaper multi-level cell (MLC) drives. Besides millions of drive-hours, the group also studied 10 different models of enterprise and consumer SSDs, from three different memory types - MLC NAND, SLC NAND, and eMLC NAND.

The study also shows that RBER (raw bit error rate) is a more dependable measure of reliability than UBER (uncorrectable bit error rate) mentioned in drive specs or datasheets. RBER increases slower than expected from wearout, and isn't correlated with UBER. However, the measured/real-world UBER is higher for SSDs than HDDs. This means that while SSDs are less likely to "fail" than HDDs, they're more likely to lose portions of their data. Keep your SSDs regularly imaged. Age, rather than usage, affects reliability of SSDs. A disturbing 30-80% of SSDs in the study developed at least one bad-block, and 2-7% of the SSDs developed at least one bad chip, within the first 4 years of deployment.



View at TechPowerUp Main Site

 

[Jetster]

I was just reading that

 

[FordGT90Concept]

So HDDs are still king for archival purposes. Why am I not surprised? At least there's imperical evidence of the fact now.

 

[OneMoar]

So HDDs are still king for archival purposes. Why am I not surprised? At least there's empirical evidence of the fact now.

not really if you are aren't writing to it and the drive is 'cold' e.g powered down then SSDs are the better option as they aren't as sensitive to thermo-magnetic disruptions
all this article says is that expensive SSDs are more reliable then cheap SSDS go figure ...

 

[R-T-B]

all this article says is that expensive SSDs are more reliable then cheap SSDS go figure ...

It actually says:

One of the study's biggest finding is that high-end (read: enterprise) SSDs with single-level cell (SLC) NAND flash memory are no more reliable than cheaper multi-level cell (MLC) drives.

and

This means what while SSDs are less likely to "fail" than HDDs, they're more likely to lose portions of their data.

So no.

 

[OneMoar]

It actually says:
and
So no.

so nothing new its a silly study
thats the problem with these kind of 'studys` they give the appearance of providing new information but all they really do is spin it in a different way
(read: enterprise) SSDs with single-level cell (SLC) NAND flash memory are no more reliable than cheaper multi-level cell (MLC) drives. which says the expensive SLC cells vs Cheap consumer sells
its spin doctoring nothing more
we have known for ages that SLC drives while having a total lower failure rate are more prone to data loss when they do fail vs MLC drives that have a higher total failure rate but have a lower chance of data loss when a cell fails

 

[R-T-B]

so nothing new its a silly study
thats the problem with these kind of 'studys` they give the appearance of providing new information but all they really do is spin it in a different way
(read: enterprise) SSDs with single-level cell (SLC) NAND flash memory are no more reliable than cheaper multi-level cell (MLC) drives. which says the expensive SLC cells vs Cheap consumer sells
its spin doctoring nothing more

Hmmm.... Whom should I trust... google, who has a stake in protecting data, or OneMoar, from the internet?

Please provide evidence of spin-doctoring. Otherwise, my choice is obvious.

 

[OneMoar]

Hmmm.... Whom should I trust... google, who has a stake in protecting data, or OneMoar, from the internet?

Please provide evidence of spin-doctoring. Otherwise, my choice is obvious.

http://0b4af6cdc2f0c5998459-c0245c5...rackcdn.com/23105-fast16-papers-schroeder.pdf
read the paper for your self silly (look a the uncorrectable errors on SLC vs MLC)
and the final nail in this silly little bit of retarded ness is that studies such as this become less accurate the longer they are run

solid state tech has come a long long LONNNNNNNG way in 6 years so the entire study is moot due to changes in SSD tech over time you can't make a apples to apples comparison if you are mixing 3,5 and 6 year old apples into a pie

 

[TheinsanegamerN]

not really if you are aren't writing to it and the drive is 'cold' e.g powered down then SSDs are the better option as they aren't as sensitive to thermo-magnetic disruptions
all this article says is that expensive SSDs are more reliable then cheap SSDS go figure ...

In a cold state, a ssd will only have data retention for 2-3 months. A hard drive, at minimum, can go 7-12 months, typically it is common to go years and still have retained data integrity. That being said, tape backups are really the only good "true" cold storage solution.

Also, the article is about how expensive SSDs based on SLC are not more reliable than cheaper MLC SSDs.

http://0b4af6cdc2f0c5998459-c0245c5...rackcdn.com/23105-fast16-papers-schroeder.pdf
read the paper for your self silly (look a the uncorrectable errors on SLC vs MLC)
and the final nail in this silly little bit of retarded ness is that studies such as this become less accurate the longer they are run

solid state tech has come a long long LONNNNNNNG way in 6 years so the entire study is moot due to changes in SSD tech over time you can't make a apples to apples comparison if you are mixing 5 3 and 6 year old apples into a pie

OK, I have no idea what you are on about. This study just came out, and now you're sitting here saying it's unreliable because its been run too long, when it JUST CAME OUT? Also, when was the last time SLC tech changed much? MLC has been the name of the game for awhile, SLC has been largely stagnant with exception to capacity.

 

[OneMoar]

In a cold state, a ssd will only have data retention for 2-3 months. A hard drive, at minimum, can go 7-12 months, typically it is common to go years and still have retained data integrity. That being said, tape backups are really the only good "true" cold storage solution.

Also, the article is about how expensive SSDs based on SLC are not more reliable than cheaper MLC SSDs.

I should have been more clear by powered-down I didn't mean offlined I ment to say sata power state such as deep sleep where as the drive is physically powered but, in a high pstate

 

[R-T-B]

http://0b4af6cdc2f0c5998459-c0245c5...rackcdn.com/23105-fast16-papers-schroeder.pdf
read the paper for your self silly (look a the uncorrectable errors on SLC vs MLC)

They are going by complete failure aparently, and saying RBER is a more reliable measure of that than UBER. That may be true, and yes, would not be what I or most would consider reliability as SSD failure itself is uncommon. (The study admits this)

and the final nail in this silly little bit of retarded ness is that studies such as this become less accurate the longer they are run

solid state tech has come a long long LONNNNNNNG way in 6 years so the entire study is moot due to changes in SSD tech over time you can't make a apples to apples comparison if you are mixing 5 3 and 6 year old apples into a pie

They appear to be age grouped to keep apples-to-apples valid.

 

[OneMoar]

They are going by complete failure aparently, and saying RBER is a more reliable measure of that than UBER. That may be true, and yes, would not be what I or most would consider reliability as SSD failure itself is uncommon.



They appear to be age grouped to keep apples-to-apples valid.

they are grouped but they make no mention of when the drive was put in service and when it was taken out
they also are using program erase and a metric which frankly is silly as you don't normally see that many PE events in normal operation
doing a actual erase on a cell is a costly maneuver thats hard on the drive its why we have TRIM

 

[R-T-B]

they are grouped but they make no mention of when the drive was put in service and when it was taken out

Fair enough. It was a quick skim and I assumed that the grouping would be logically based on age, I did not validate that.

 

[OneMoar]

Fair enough. It was a quick skim and I assumed that the grouping would be logically based on age, I did not validate that.

they also don't specify what controler and what brand of NAND nor the fab process it was built on
early SLC was crap expensive,little redundancy for bad cells,slow when you needed todo multiple operations on a cell ... which is why we ended up with MLC

 

[OneMoar]

if you really wanna know what SSD can take a beating then THIS is what you wanna read
http://techreport.com/review/26058/the-ssd-endurance-experiment-data-retention-after-600tb
and
http://techreport.com/review/27909/the-ssd-endurance-experiment-theyre-all-dead

 

[FordGT90Concept]

In a cold state, a ssd will only have data retention for 2-3 months. A hard drive, at minimum, can go 7-12 months, typically it is common to go years and still have retained data integrity.

I was going to object to the 7-12 month figure but before I said as much, I just plugged in a 1997 Pentium II Gateway 2000 machine. It powered on but refused to give me any video. It might not like my KVM or the fact that keyboard is plugged in via USB instead of PS/2 so I gave up. I then tried my 2000 Compaq Presario with an Athlon processor. It shows the Compaq logo but it sounds like it just keeps POSTing over and over again. Both systems haven't been plugged in for years.

Without further testing, I don't know what specifically broke in the systems but at this point, the above statement appears to be true. I expected both systems to run without a fuss.

I'm debating about just pulling the HDDs out of them and erasing them. I'm pretty sure I erased them before I used them for old games so not really anything of value on them. It would tell me if the HDDs are still in good shape or not though.

 

[OneMoar]

I was going to object to the 7-12 month figure but before I said as much, I just plugged in a 1997 Pentium II Gateway 2000 machine. It powered on but refused to give me any video. It might not like my KVM or the fact that keyboard is plugged in via USB instead of PS/2 so I gave up. I then tried my 2000 Compaq Presario with an Athlon processor. It shows the Compaq logo but it sounds like it just keeps POSTing over and over again. Both systems haven't been plugged in for years.

Without further testing, I don't know what specifically broke in the systems but at this point, the above statement appears to be true. I expected both systems to run without a fuss.

I'm debating about just pulling the HDDs out of them and erasing them. I'm pretty sure I erased them before I used them for old games so not really anything of value on them. It would tell me if the HDDs are still in good shape or not though.

dead cmos battery lol

 

[FordGT90Concept]

It should be able to POST/boot without CMOS battery. Both systems should be able to run off of factory defaults. I suppose if I were to fully troubleshoot them, that should be the first thing I check.

 

[OneMoar]

It should be able to POST/boot without CMOS battery. Both systems should be able to run off of factory defaults. I suppose if I were to fully troubleshoot them, that should be the first thing I check.

not with old systems
change battery clear cmos and it should post

 

[Scrizz]

It should be able to POST/boot without CMOS battery. Both systems should be able to run off of factory defaults. I suppose if I were to fully troubleshoot them, that should be the first thing I check.

The no video issue on the Pent2 is totally not related to the HDD.

 

[Bayfront Benny]

If you image your drive regularly in order to avoid uncorrectable data loss, how would you know that you do not have damaged data in the first place? And if you don't know - you would be passing the damaged data on to your image as well. This of course would hold true for all hard drives not just ssd's.

 

[PP Mguire]

In a cold state, a ssd will only have data retention for 2-3 months. A hard drive, at minimum, can go 7-12 months, typically it is common to go years and still have retained data integrity. That being said, tape backups are really the only good "true" cold storage solution.

Also, the article is about how expensive SSDs based on SLC are not more reliable than cheaper MLC SSDs.



OK, I have no idea what you are on about. This study just came out, and now you're sitting here saying it's unreliable because its been run too long, when it JUST CAME OUT? Also, when was the last time SLC tech changed much? MLC has been the name of the game for awhile, SLC has been largely stagnant with exception to capacity.

Unlike Ford, my stuff actually worked. I've had a 1TB drive offline sitting in a box tagging along with me for about 5 years that still has a working useable Vista install. I've had my Corsair Neutron GTX 120GB sitting in my backpack since September and when I plugged it in proceeded to boot up Windows 10. I'd say this is hog wash at best.

 

[Frick]

I recently went through a bunch of old laptop HDDs. One of them is a 3GB drive from the late 90s iirc,and it worked flawlessly.

 

[yogurt_21]

I recently went through a bunch of old laptop HDDs. One of them is a 3GB drive from the late 90s iirc,and it worked flawlessly.

have a 245MB drive from a 486 that still powers up and works great. Has windows 3.1 on it and it is still the fastest booting GUI + HD combo I've ever had.

 

[FordGT90Concept]

If the drives still work, the Compaq has a 40 GB and the Gateway 2000 has 1-3 GB. I suspect they're fine but I can't be arsed to troubleshoot systems that really have no use today.

The Compaq did not age well. It's louder than my server!