4TB drives cheaper in USB?

[DaPlumber]

Has anyone else noticed that you can get the "consumer" 4TB SATA drives cheaper than bare drives if you're willing to shuck it out of a USB3 shell? The sale/discounted price seems to be hovering around $130 in multiple channels and across brands: So far I've seem WD, Samsung, Toshiba, and Seagate.

Is this just a temporary supply/demand aberration or is there something else at work here? I can't imagine that the manufacturers sell so many more USB3 enclosed drives that it offsets the additional costs?

Is it also my imagination that the manufacturers are trying their damnedest to make the drives difficult to extract without destroying the shell too?

Weird.

 

[cyberjock]

Most USB drives have a very short warranty, some are just 6 months. Also some drives won't work "the same" if you break them out of the USB enclosure. Still others haven't worked at all for me outside of the enclosure.

So you might think you are saving yourself some money, but you might not be if your drive fails prematurely.

 

[DaPlumber]

Sound advice. However the 4TB models I've seen recently have the same 1 year warranty as the "bare" consumer drives. Obviously the warranty is void after shucking, but at that price point I'm not particularly fussed. The reliability should be very similar and while not the same as the NAS specific drives, not every application calls for that. I know that Seagate in particular went though a phase where their USB drives ran different firmware, but this 4TB/USB3 generation seem to not be doing that (reading reviews of dissections and shuckers.) I can personally verify that the WD model currently on offer that I received had a completely standard Green drive in it.

BTW Kids: Drive manufacturers track which serial numbers go into their own shells, so the warranty is N/A if you tried to claim it on the bare drive.

Anyhoo, I'm far more interested in the Economic "whys" of this than suggesting someone use a bunch of shucked drives to save money as their primary pool. Backup pool? Maybe, but goblin with your eyes wide open as to the compromises you're making. (Plenty of material in these forums about that.)

Now for some speculation: since manufacturers are bringing "mainstream" 5TB and 6TB drives on stream, could it be they're trying to flush stock of 4TB consumer drives out via discounted USB3 models?

 

[cyberjock]

The why is probably because "the business model says so". Let's face it though. If you are buying USB3 drives (and most people don't remove the drive from the shell) they aren't likely to do significant data transfer (USB sucks!). Most people use it as a backup drive or just throw a bunch of data on it in bulk and then access it once a day or less. With that little of a usage pattern I'm sure the failure rates are going to be fairly low.

I seriously doubt that manufacturers are trying to flush their stock of 4TB drives. Manufacturers typically can't keep a large inventory of spare drives (especially of their largest size) because they can't manufacture them at a fast enough rate to maintain a large inventory. I haven't been able to confirm this for 100% certainty, but my understanding is that the 5 and 6TB drives do not have a higher density per square inch than 4TB drives and they skirt around the problem with more platters. So the worst the manufacturers could be accused of is thinning their margins. At least, in my opinion. ;)

 

[DaPlumber]

I think you're correct about the business model/channels. Maybe they overestimated demand for 4TB consumer drives 6-12 months ago? More like trying to flush the channel than their own on-hand stock (which they don't have). You're quite correct the game isn't played that way any more, I should have been clearer.

Areal density isn't everything, new platter/motor/servo tech. can lead to increases too. Hitachi's Helium approach is one example, although I don't think that one's going to be cost effective, but I could be wrong. The first 6TB drives are "exotics" with more current tech. platters certainly, but the first "shingle" platters are coming too.

I concur on duty-cycle. These drives to me are great candidates for a second pool for backup or media storage under FreeNAS and given FreeBSD's apparent disdain for USB mass storage, I'd shuck 'em and use SATA. I will say that under Solaris I've used USB drives with ZFS with none of the issues I've encountered under FreeNAS/FreeBSD and with no issues under Linux and OS X with their native formats.

But, since this is a FreeNAS forum: "USB BAD!" :p

Not to be a conspiracy nut and wrenching back on original topic, but I still think there's something odd going on as it just seems to be the 4TB drives that are being discounted this much with USB3 shells. It could be that one manufacturer over-produced 6 months ago and the rest are forced to match. Then again drive manufacturers are an interesting cabal... ;)

Any other drive shuckers or potential drive shuckers out there? ;)

 

[cyberjock]

The USB problem is one that is OS agnostic. Regardless of what OS you use the underlying problems with USB have to do with the technology and not the drivers or anything else that anyone has control of. It's strictly a silicon problem. It's entirely possible to build a server out of USB drives and have it work for a very long time. That's not the problem though. The problem is that ZFS makes assumptions about how to flush disk caches, that data it sends to the "SATA" controller really does get to the disk, and things like that. USB effectively breaks that in ways that aren't fixable because it's a silicon problem. So it can work great for a long time. But the odds of you hitting that one bump in the road that will ruin everything is pretty high. The solution is to just never use USB. And lets face it, how many enterprises buy 50 USB drives and make a server? None. The reason is simple, it's not a proper use of the technology. ;)

I think the real problem is that these companies weren't expecting the mighty cyberjock to hurt their USB business by telling the world that they shouldn't make servers with them. Business was good until the truth came out. They were unprepared for how much pull I have in their markets and they've had to succumb to me! /sarcasm

 

[gpsguy]

IIRC, the Blackbaze folks have been known to shuck the USB drives and use them in the servers.


Sent from my phone

 

[Ericloewe]

It's odd to see such pricing. Last time I bought external drives, I bought two apparently identical WD 2TB ones. Turns out one was a more recent revision and had a USB cable with rounded edges and an EARX Green disk inside (which is coincidentally sitting on my desk, barely capable of reporting its SMART data...) while the other one came with an EARS disk. My interpretation is that they don't care about flushing out old stock in a hurry - they just sell it as being identical to the newer units.

IIRC, the Blackbaze folks have been known to shuck the USB drives and use them in the servers.


Sent from my phone

I remember such a thing, too. Must've been a hell of a price difference to offset paying some guys to spend their day opening external enclosures.

 

[DaPlumber]

@Cyberdog: No, not all USB mass driver stacks are created equal. ;) I have personally done horrible things to Solaris connected USB drives on multiple occasions. The FreeBSD crowd don't seem to consider the USB mass-storage (and other) driver(s) a priority (I haven't tried 10 yet, full disclosure). <Shrug>

I see a 4TB WD Red drive from NewEgg crossed my inbox a few days ago for $159 (with code), so I guess it's 4TB price drops all around and the USB drives were just first?

 

[cyberjock]

@Cyberdog: No, not all USB mass driver stacks are created equal. ;) I have personally done horrible things to Solaris connected USB drives on multiple occasions. The FreeBSD crowd don't seem to consider the USB mass-storage (and other) driver(s) a priority (I haven't tried 10 yet, full disclosure). <Shrug>

Our problems aren't with the USB drivers. Our problem is with the USB protocol, how USB to SATA bridges function etc. So yes, I stand by my comment that it is OS agnostic because our problem is NOT with the USB drivers. Yes, USB3 drivers need work on FreeBSD, but that's not our problem.

When the hardware itself is the problem and it's not drivers, it's OS agnostic. ;)

 

[DaPlumber]

@Cyberdog: Eh, let's give the deceased equine one more smack and say that the problem may be OS agnostic, but how the driver writers handle it is not. For all the other things I have issue with in various incarnations of Solaris: whoever wrote the USB stack seems to have an almost paranoid distrust of USB bridges and a fondness for watchdog timers. That's rather to your point? ;) At least that's the impression I recall I got on reading log messages after abusing the stack many moons ago.

For the record: USB3 on FreeNAS 9.2.1.5-RELEASE was no better or worse than USB2 from my (admittedly limited) testing. I understand why it's disabled by default when there's a specific hardware set out there that can blow up... :D See paranoid driver coding above...:p

One last pet peeve kids: NEVER trust a USB storage device that is not adequately powered independent of the USB port. NEVER EVER

Now that 5TB and 6TB drives that aren't complete science projects seem to be rolling out it looks like all 4TB drives have taken or will shortly take a haircut. Now there really is no excuse for not using "NAS rated" 4TB drives.

 

[cyberjock]

You aren't arguing the points I'm arguing. This isn't about USB2 or USB3. This is about how USB responds to problems. My good example: If a USB hard drive encounters a single bad sector there's a >90% chance that the drive *will* be disconnected from the USB interface. That means your pool just lost a disk because of a single bad sector. So that ultra-redundant zpool can be ruined forever because of just a few bad sectors or bad communication errors.

There's a ton more examples, but I really don't want to get into it. This isn't a discussion you are going to win. This forum has plenty of "examples" of how bad USB is long-term and I don't need to prove my point. The forums can do that for me.

So no, the tech is not fit for long-term use, will never be fit for long term use, and anyone that uses it will find it works great until the second it doesn't. And when it doesn't it *really* doesn't. And since those problems can come without warning and despite all of your attempts to pre-plan for the eventuality and not save you there's serious uncorrectable risk involved with USB.

I don't even argue the USB3 drivers or USB2 driver maturity because that's something that can (and most likely will) be fixed someday. USB2 works very well for USB chipsets that are made by manufacturers that have decided to invest time in coding good USB drivers. The rest of the manufacturers that choose not to participate in all of the FreeBSD goodness that we love here will likely never work. That's not exactly our problem directly.

But the fact that USB interfaces disconnect so easily cannot be fixed without redesigning USB's basic silicon. And redesigning USB's basic silicon effectively means "not use USB" as the changes that would need to be made to the communications channels and would not be backwards compatible with prior versions of USB. And part of the marching on with USB revisions and specifications is that things must be backwards compatible.

 

[DaPlumber]

I'm not trying to "win" anything. Down doggy! :p

What I'm obviously not doing well is communicating, my bad, so I'll try again: What I'm trying to say is that different OS driver stacks respond differently to the same sucky hardware and/or hardware architecture issues. For a trivial example: If there is a bus error or an I/O timeout OS "A" may immediately propagate that back up the stack as a fatal I/O error and may not have a choice because of the hardware involved. OS "B" may back off reset the bus and retry the I/O and if it succeeds report a soft error or warning but not propagate a fatal I/O. OS "C" may even let you choose the timeouts, the fatality of the error and even give you the ability to unload "hung" bits of driver and if the hardware supports it power cycle the offending bits of the hardware. This is not peculiar to USB, or personal/consumer computing. None of the approaches above is "better" than the other, they just have different design criteria and priorities. The shame is on me for making assumptions about FreeBSD that just because it's running ZFS the driver stack behaves the same way as Solaris because they share a common ancestry. I also forget that when I say "use USB" in this context I mean use each device on a separate USB bus/hub which minimizes SOME of the issues. I'll willingly accept the forehead slap for those two.

Full disclosure: Many moons ago I reviewed and tested low-level I/O software including for USB interfaces. I fully agree there are plenty of things about USB that can't be "fixed" that are stupid and/or broken (See recent security hysteria, It's been known for years that USB devices can "impersonate" other devices, someone finally got around to talking about weaponizing it in public.) All you can do is try and work around them IF it's deemed to be "worth it".;)

tl;dr: Kids, don't use USB devices with FreeNAS unless you're prepared to lose the pool. YMMV elsewhere, but that still doesn't make it a good idea.