Friday, April 24th 2020

Western Digital Spells Out Which Specific HDD Models Use SMR

Western Digital on Thursday (23/04) released an updated statement about the company's stand on SMR (shingled magnetic recording) being used on certain internal hard drives, including certain client-segment WD Red models recommended by the company for NAS applications, without proper disclosure on the product's marketing materials or data-sheets. It's surprising to note that SMR is being used not in some of the higher-capacity models (8 TB or higher) as previously thought, but rather lower-capacity ones, 6 TB or lower, including a 2 TB 3.5-inch drive, and a 1 TB 2.5-incher.

Perhaps SMR is being used in these lower-capacity drives to reduce the number of platters, by cramming in more data per recording surface. As on 22nd April (when the list was internally compiled by Western Digital), there are 9 client-segment hard drives that use SMR, four of which are from the company's WD Red family, two each from the WD Blue 3.5-inch and WD Blue 2.5-inch families; and one from the 2.5-inch WD Black family. Among the WD Red series drives with SMR are the 2 TB WD20EFAX, the 3 TB WD30EFAX, the 4 TB WD40EFAX, and the 6 TB WD60EFAX.
The undisclosed SMR controversy blew up on Western Digital's face last week, when a Blocks & Files report chronicled incidents of several WD Red series HDDs being rejected in RAID volumes during their resilvering process. Over the course, it was learned that even Seagate and Toshiba were selling HDDs with undisclosed SMR, though none marketed as "NAS optimized" ones.

Western Digital earlier this week posted its first response to the controversy, by admitting that some WD Red HDDs use drive-managed SMR (DM-SMR), an internal SMR implementation that's transparent to the host, and pointing NAS users to the company's WD Red Pro, WD Gold or even enterprise-grade Ultrastar product lines. SMR is a physical-layer recording technique that overlaps write tracks over each other to achieve greater data density at the high cost of random write performance, which makes them effectively unfit for serious RAID applications (in turn NAS applications that use some form of internal RAID). Western Digital also provided a general guideline as to which class of its client-segment hard drives are likely to feature SMR.
Source: Western Digital
Add your own comment

44 Comments on Western Digital Spells Out Which Specific HDD Models Use SMR

#1
notb
"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR.
Please stop repeating this.
Posted on Reply
#2
framebuffer
it wasn't that difficult after all, right WD?
Posted on Reply
#3
wiak
notb"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR.
Please stop repeating this.
how was everyone supposed to know that?
i got 3x WD60EFRX & 1x WD60EFAX

the EFAX was a replacement for a failed EFRX
Posted on Reply
#4
notb
wiakhow was everyone supposed to know that?
Probably multiple ways, like checking the NAS compatibility lists etc.

But I'm not sure why you're asking me that.
We know SMR drives work in NAS and RAID. We know that they have different characteristics than CMR. That's it.
Author of this news is repeating an incorrect information that people start to believe in. That's the problem.

Also, the SMR/CMR issue has been discussed for months now.
voodooFXit wasn't that difficult after all, right WD?
Hmm. Let's ask Seagate and Toshiba. Where are their statements? :)
Posted on Reply
#5
Tomorrow
notbProbably multiple ways, like checking the NAS compatibility lists etc.
That's assuming everyone using RAID on these drives used them in NAS in the first place. That's like saying to check the motherboard manufacturers manual for CPU specs.
Posted on Reply
#6
wiak
TomorrowThat's assuming everyone using RAID on these drives used them in NAS in the first place. That's like saying to check the motherboard manufacturers manual for CPU specs.
Tomorrow's logic Today XD
Posted on Reply
#7
notb
TomorrowThat's assuming everyone using RAID on these drives used them in NAS in the first place. That's like saying to check the motherboard manufacturers manual for CPU specs.
You're absolutely correct.

Then again, that leaves us with people who use RAID on DIY desktops (or at least added HDDs to an OEM desktop on their own, i.e. not following the OEM recommendations).
Well, DIY group is always vulnerable to this kind of problems. That's why everyone else prefer to pay OEMs to take care of some matters. :)

Still, it's a little unfair that PC websites focus on WD, when the other 2 major HDD makers did almost the exact same thing.
Yes, WD is the only company that marketed consumer SMR drives for NAS - Toshiba and Seagate only use SMR for desktop and archive drives.
That was a mistake and they should fix it. This statement is a temporary solution - they have to correct product datasheets and ideally describe the consequences and how to safely use SMR drives in RAID.

But at that point the overall story given by PC websites becomes incoherent, because:
1) many news authors focus on SMR performance, which obviously affects desktop drives,
2) RAID in a PC doesn't have to be made up from NAS drives.
Posted on Reply
#8
plumflorin
@notb, stop trying to look smart and cocky! Not everybody know that some Sata drives doesn't work with other sata drives. In the past they were all compatible. They brake that compatibility so if they wanted to be fair, hdd manufacturers should've put a label that explicitely says that SMR drives are not compatible with other sata drives. From my point of view, I hope a class action lawsuit will be initiated to make them pay for false advertising.
Posted on Reply
#9
notb
plumflorin@notb, stop trying to look smart and cocky! Not everybody know that some Sata drives doesn't work with other sata drives. In the past they were all compatible.
Consumer PC SMR drives have been around at least since 2016.
Server and consumer backup/archive drives were launched in 2014 or maybe even earlier.

Please look it up before you decide to sue me. :)
Posted on Reply
#10
wahdangun
notb"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR.
Please stop repeating this.
Wtf, the problem is not about smr and never about that, the problem was WD changed their red drive specification and making it not compatible to use, that's definition of bait and switch, and they can be sued by that
Posted on Reply
#11
newtekie1
Semi-Retired Folder
notb"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR.
Please stop repeating this.
This is incorrect. They work fine initially. But the minute you have to do a rebuild or OCE/ORLM the SMR becomes a major problem. This is particularly true on rebuilds. The reason being the SMR makes these operations take insane amounts of time, drastically increasing the chances of an additional drive failure happening during the process. When an array is in the critical state you want to get it out of that state as fast as possible, using drives that are known to be extremely slow in the processes that RAID uses during a rebuild(random writes across the entire drive) is not something you should be doing.

It may seem like if you just match all CMR or all SMR drive, all will be dandy. But a CMR array rebuild will take hours, maybe a couple days depending on the size of the drive. While a SMR rebuild is going to take weeks. SMR drives are not suited for RAID use because of this.
notbHmm. Let's ask Seagate and Toshiba. Where are their statements? :)
Seagate's statement was that they would never use SMR in a product designed for RAID/NAS use.
notbThen again, that leaves us with people who use RAID on DIY desktops (or at least added HDDs to an OEM desktop on their own, i.e. not following the OEM recommendations).
Well, DIY group is always vulnerable to this kind of problems. That's why everyone else prefer to pay OEMs to take care of some matters. :)
You think those are the only two options? Either they are using a NAS or they are a DIY doing RAID in a desktop? Sorry, nope.

Professionals aren't all using NAS devices. We're using Servers with RAID, and we're buying these Red drives.
notbConsumer PC SMR drives have been around at least since 2016.
Server and consumer backup/archive drives were launched in 2014 or maybe even earlier.

Please look it up before you decide to sue me. :)
No one is saying SMR is new. He/we aren't talking about problems with backup/archive drives, so I don't know why you are even bringing them up. Our problems are with the Red drives. His statement is about mixing SMR drives with CMR drives in a RAID configuration like they are marketed for.
Posted on Reply
#14
redwrx02
I consider myself very lucky that I chose the Seagate Ironwolf NAS drives over the WD Reds! It seems the Red Pros are also a premium over the corresponding Ironwolf models.
Posted on Reply
#15
windwhirl
notbHmm. Let's ask Seagate and Toshiba. Where are their statements? :)
They both told Blocks and Files about it and gave a list of drives that are not using SMR, but I'm sort of pissed at all three of them for not saying anything about SMR in their goddamn datasheets.

It couldn't be that hard to add an extra row for the technology used...
wiakthe EFAX was a replacement for a failed EFRX
Sooo, they gave you a drive with different specs than the one you had without telling you anything.

A little honesty is all we ask of them, and they can't even do that.
Posted on Reply
#16
Zareek
I'm not a fan of how WD handled this. Something that affects drive performance radically should be presented openly up front. Based on the chart, if it is a reasonably priced WD drive, it has SMR. Guess I'm not using WD Reds for SO/HO NAS setups anymore. Too bad, they have been really reliable!
Posted on Reply
#17
notb
newtekie1It may seem like if you just match all CMR or all SMR drive, all will be dandy. But a CMR array rebuild will take hours, maybe a couple days depending on the size of the drive. While a SMR rebuild is going to take weeks. SMR drives are not suited for RAID use because of this.
SMR rebuild will realistically last few times longer than on CMR of the same capacity. Absolutely true.

Well, that's how RAID rebuilds are performed. Maybe as SMR drives become more popular, we'll get a more suitable algorithm for rebuilding. SMR is really modern tech after all. But that's something we should demand from controller / server / NAS makers, not from drive manufacturer.

Because in turn buyers expect HDD makers to increase capacity. That's why we have SMR. That's why we'll see other recording methods in next few years. There's no going back to all-CMR universe.

And while I agree with you, this is still a performance aspect.
Something similar could be said about SSD vs CMR HDD. Rebuild takes few times longer. It's just a matter of what you're used to.

Saying that long rebuilds increase a chance of failure is... mathematically true. And not easy to respond to.
If you don't feel comfortable with rebuilds, don't do them. Restore from a backup. It's usually faster anyway.
Seagate's statement was that they would never use SMR in a product designed for RAID/NAS use.
And yet problems with Seagate SMR drives were reported long before WD Red SMR drives were even launched. People wanted to run drives from the Archive lineup in RAID.
Because, as someone already said, RAIDs aren't exclusive to NAS.

Yes, I agree most people run NAS with RAID, so it should be explicitly said what the SMR consequences are. But I expect this information for all SMR drives, from all manufacturers.
Professionals aren't all using NAS devices. We're using Servers with RAID, and we're buying these Red drives.
No offense, but was this meant to impress me or what? It really looks silly. :D

Don't buy consumer Red drives for enterprise applications. There's a product line for that.
Posted on Reply
#18
BSim500
notb"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR. Please stop repeating this.
That depends entirely on usage. You may get away with SMR with light use of one person copying a few files over as a backup onto a nice clean drive with the NAS doing nothing else, but if Person A is streaming a video from a NAS whilst person B copies some files over, the difference between SMR vs CMR effectively forcing a multiplication of required writes via having to rewrite every other files on surrounding tracks (especially on a fragmented drive) could easily make the difference between person A experiencing an unwatchable stutter-fest or not, at which point yes SMR can absolutely be woefully unsuitable for multi-user NAS even if the drives are not mixed. The main issue though is that given an informed choice, most people wouldn't voluntarily buy SMR over CMR them for 3.5" drives for desktop / NAS use for the sake of saving $5 (if that), and the only way WD sold them was via a dishonest "bait & switch".
Posted on Reply
#19
Tomorrow
redwrx02I consider myself very lucky that I chose the Seagate Ironwolf NAS drives over the WD Reds! It seems the Red Pros are also a premium over the corresponding Ironwolf models.
Not only that but for some idiotic reason Red Pro's also cost more than Ultrastar. Despite Ultrastar having superior specifications most of the time. For example 14TB Red Pro vs Ultrastar DC HC 530.
Posted on Reply
#20
newtekie1
Semi-Retired Folder
windwhirlThey both told Blocks and Files about it and gave a list of drives that are not using SMR, but I'm sort of pissed at all three of them for not saying anything about SMR in their goddamn datasheets.

It couldn't be that hard to add an extra row for the technology used...
They aren't using SMR in drives where SMR could be a problem though, so Seagate and Toshiba really aren't doing anything wrong. SMR has been in use for years, as was pointed out. This is nothing new, we all knew it was in use.

Putting it in a product intended for RAID is where the issue comes up.
windwhirlSooo, they gave you a drive with different specs than the one you had without telling you anything.

A little honesty is all we ask of them, and they can't even do that.
I wouldn't be surprised if the RMA department didn't even know there was a difference between the drives. Or at least not as big of a difference as there really is. It seems WD wasn't even properly informing their own staff that SMR was used in any of their drives.
notbSMR rebuild will realistically last few times longer than on CMR of the same capacity. Absolutely true.

Well, that's how RAID rebuilds are performed. Maybe as SMR drives become more popular, we'll get a more suitable algorithm for rebuilding.
If that happens, then by all means, re-launch them as RAID/NAS drives. But until then, the fact is SMR drives are not suitable for NAS/RAID use, as stated in the original article.
notbSMR is really modern tech after all.
It's not new. It's been in use since 2013. That's ancient by computer terms.
notbBut that's something we should demand from controller / server / NAS makers, not from drive manufacturer.
No, it is absolutely something we should demand from the drive manufacturers. It is their product's performance that is the issue, not the NAS/RAID controllers. The rebuild time is almost always dictated by the speed at which the data can be written to the drives. Introducing an extremely slow drive kills rebuild performance. There isn't some magical algorithm that RAID developers can use where they can rebuild an array by some how not writing data to the new drive. It's just not possible.

On top of that, these drives report to the system the same as a CMR drive. It should be pointed out that there are enterprise SMR drives, but they actually have a method in place to report to the controller/system that they are SMR and the system handled those drive differently. In the case of the WD Red drives, they don't do that, because they are using DM-SMR. So they appear to the system as a CMR drive, and the controller on the drive itself handles all the SMR translation.
notbBecause in turn buyers expect HDD makers to increase capacity. That's why we have SMR. That's why we'll see other recording methods in next few years. There's no going back to all-CMR universe.
This statement goes totally against the facts. Yes, SMR can be used to increase capaity. And if we were seeing SMR used in super large capabity Red drives, I'd almost agree with your argument. But WD only used SMR in the lowest capacity Red drives. Explain that! If SMR is about getting higher capacities, why is it only used on the 2, 3, 4, and 6TB sizes and not the 8, 10, 12, 14TB drives. It's 100% not necessary for drives 14TB and smaller, we know this because there are 14TB CMR drives on the market. So your argument doesn't apply.
notbAnd while I agree with you, this is still a performance aspect.
Something similar could be said about SSD vs CMR HDD. Rebuild takes few times longer. It's just a matter of what you're used to.
Sure, an SSD array would rebuild significantly faster. But very large capacity SSD arrays are not financially practical for most situations. So we use HDDs. If SSDs were an option, trust me, we'd all be using them. I can come up with tons of unlikely scenarios too if you want, but lets stick to reality, shall we?
notbSaying that long rebuilds increase a chance of failure is... mathematically true. And not easy to respond to.
If you don't feel comfortable with rebuilds, don't do them. Restore from a backup. It's usually faster anyway.
I have a feeling you've never really had experience with RAID from all your statements. Doing this would require the storage system to be taken offline and not be accessible during the restore. Plus, re-creating the array would require an initialization before the array become redundant, which I'll tell you right now takes a hell of a lot longer than just doing a rebuild. So, no, just restoring from a backup is not usually faster.
notbAnd yet problems with Seagate SMR drives were reported long before WD Red SMR drives were even launched. People wanted to run drives from the Archive lineup in RAID.
Because, as someone already said, RAIDs aren't exclusive to NAS.
Yep, and when you run drives outside of their intended purpose, issues happen. Again, you're focussed on your idea that we are complaining about SMR. The issue isn't SMR, it's SMR put in drives specifically marketed for RAID/NAS use.

I was one of those people that took archive drives and ran them in RAID5. I had 3 5TB archive drives. It's exactly how I know the problems that SMR drives cause in RAID. It ran just fine for a couple years, until one of the drives died. I was actually never able to get it to rebuild. I put a replace drive in that was the same model as the others, and when the new drive reached the end of its CMR cache, the controller would see the huge write performance drop, assume the drive was bad and kick it back out of the array. I finally had to give up on it, and I'd never use SMR in RAID again. It's just not worth it.
notbYes, I agree most people run NAS with RAID, so it should be explicitly said what the SMR consequences are. But I expect this information for all SMR drives, from all manufacturers.
It'd be nice if every manufactuere diclosed it. But it's only required on drives marketed for NAS/RAID use. And that's the distinction here and why WD is in the wrong while the other manufacturers aren't.
notbNo offense, but was this meant to impress me or what? It really looks silly. :D

Don't buy consumer Red drives for enterprise applications. There's a product line for that.
No, it's meant to inform you that while you might be an armchair quarterback that's obviously never even experienced a RAID array in your life, I do it for a living.

And while there are enterprise grade drives, yes, most small and medium business as well as SOHO users are buying Red drives. They are supposed to be suited for the needs of these companies. Your statement about enterprise drives is ignorant. I mean, should be be putting WD Gold drives in desktops just because they are using at a business? No. The Red drives are not only for home use, no hard drive is. The fact is, even WD says, that the Red drives are for up to 8 Drive RAID arrays for business or home use.

Directly from the WD Red Product page:
Businesses thrive on productivity and efficiency, two guiding principles built into the design of the WD Red drive. Share and backup files at the speed of your business with a WD Red drive
So it seems the "product line for that" is in fact the WD Red line. :D
Posted on Reply
#21
danbert2000
I don't care that it's on a 2.5" drive, selling a WD Black with SMR is fraud. You can't put SMR in a performance-oriented product. I can see the argument for Blue or Red, as there are better hard drive lines above them so you shouldn't expect top performance, but for that Black drive, it's downright deceptive marketing. Disgusting. I will definitely look to Toshiba and Seagate before any WD drive in the future.
notb"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR.
Please stop repeating this.
"Perfectly well," except for being one quarter of the speed, having issues reading and writing at the same time, becoming unresponsive when the cache is full.

Perfect.
Posted on Reply
#22
claes
newtekie1I have a feeling you've never really had experience with RAID from all your statements. Doing this would require the storage system to be taken offline and not be accessible during the restore. Plus, re-creating the array would require an initialization before the array become redundant, which I'll tell you right now takes a hell of a lot longer than just doing a rebuild. So, no, just restoring from a backup is not usually faster.

...

Yep, and when you run drives outside of their intended purpose, issues happen. Again, you're focussed on your idea that we are complaining about SMR. The issue isn't SMR, it's SMR put in drives specifically marketed for RAID/NAS use.

I was one of those people that took archive drives and ran them in RAID5. I had 3 5TB archive drives. It's exactly how I know the problems that SMR drives cause in RAID. It ran just fine for a couple years, until one of the drives died. I was actually never able to get it to rebuild. I put a replace drive in that was the same model as the others, and when the new drive reached the end of its CMR cache, the controller would see the huge write performance drop, assume the drive was bad and kick it back out of the array. I finally had to give up on it, and I'd never use SMR in RAID again. It's just not worth it.

...

No, it's meant to inform you that while you might be an armchair quarterback that's obviously never even experienced a RAID array in your life, I do it for a living.

And while there are enterprise grade drives, yes, most small and medium business as well as SOHO users are buying Red drives. They are supposed to be suited for the needs of these companies. Your statement about enterprise drives is ignorant. I mean, should be be putting WD Gold drives in desktops just because they are using at a business? No. The Red drives are not only for home use, no hard drive is. The fact is, even WD says, that the Red drives are for up to 8 Drive RAID arrays for business or home use.
Thanks for finally saying so — really unbearable to watch notb die on this hill for no apparent reason while talking about how reliable RAID5 is in 2020 :rolleyes:
Posted on Reply
#23
newtekie1
Semi-Retired Folder
danbert2000I don't care that it's on a 2.5" drive, selling a WD Black with SMR is fraud. You can't put SMR in a performance-oriented product. I can see the argument for Blue or Red, as there are better hard drive lines above them so you shouldn't expect top performance, but for that Black drive, it's downright deceptive marketing. Disgusting. I will definitely look to Toshiba and Seagate before any WD drive in the future.



"Perfectly well," except for being one quarter of the speed, having issues reading and writing at the same time, becoming unresponsive when the cache is full.

Perfect.
SMR, when handled correctly, won't really be noticeable to a user in a single drive configuration. The average user is not likely going to write enough data to the drive at one time to fill up the CMR cache, but I agree it is a little deceptive to call a SMR drive a "Performance" product. But there is more argument against using SMR in Red drives than Black.

IMO, SMR belongs in high capacity Blue drives and the equivalents from other manufacturers, and that's it.
Posted on Reply
#24
lexluthermiester
notb"which makes them effectively unfit for serious RAID applications "
Incorrect. SMR drives work perfectly well in RAID. The only problem is mixing SMR and CMR.
Please stop repeating this.
This is correct. Mixing the two types of magnetic recording methods on the same drive can cause slow-downs that might(under certain conditions) then cause RAID array sync errors. This applies almost exclusively to enterprise level RAID arrays and are unlikely to affect consumer level RAID setups.

As a stand alone drive, mix the two methods will only cause a slow down on a minor level that most people are completely unaware of. Such a slow down would only be revealed by a drive testing utility.
danbert2000"Perfectly well," except for being one quarter of the speed
This statement illustrates a misunderstanding of the the technology and how it is employed. SMR is NOT 25% the speed of CMR. It is closer to 94% to 97% that of CMR depending in which part of the platter is being accessed.
newtekie1Putting a mixed mode in a product intended for enterprise RAID is where the issue comes up.
Fixed that for you.
Posted on Reply
#25
claes
lexluthermiesterThis is correct. Mixing the two types of magnetic recording methods on the same drive
You mean “disks” (not methods) on the same “array” (not drive), right?
can cause slow-downs that might(under certain conditions) then cause RAID array sync errors.
Source? I think the correct answer is most of the time/high risk... Even if all of your drives are SMR, your rebuild is still going to fail once the cache fills...
This applies almost exclusively to enterprise level RAID arrays and are unlikely to affect consumer level RAID setups.
Could you please explain the difference between enterprise and consumer RAID setups as well as its relationship to this issue? I am dying to know.
As a stand alone drive, mix the two methods will only cause a slow down on a minor level that most people are completely unaware of. Such a slow down would only be revealed by a drive testing utility.
As a stand alone drive, yes, you are correct, at least under light loads.
In a networked application where multiple users are accessing the same single drive, no, you are incorrect.
In a RAID array where you need to rebuild, well, that’s just nonsense — random write performance is essential to a rebuild.
This statement illustrates a misunderstanding of the the technology and how it is employed. SMR is NOT 25% the speed of CMR. It is closer to 94% to 97% that of CMR depending in which part of the platter is being accessed.
You are talking about sequential, not random, performance — no one is disagreeing that SMR offers fine sequential performance.
Fixed that for you.
Actually, if they were “mixed mode” drives there wouldn’t be an issue at all... the drives would simply acknowledge the controller and disable SMR. The issue is that the drives aren’t capable of mixed mode, don’t point out SMR in the specs, and don’t offer instructions on how to disable SMR via some firmware utility... Please see the section on data management: en.m.wikipedia.org/wiki/Shingled_magnetic_recording
Posted on Reply
Add your own comment
Nov 21st, 2024 10:14 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts