WD RED Drives (WD10EFRX WD20EFRX WD30EFRX)

[foxbox]

Have anyone seen a straightforward answer from WD about what the Error recovery controls in their NASware really does?

Whether these drives do have TLER or not is debated, see these posts for example:
http://lars.werner.no/?p=841
http://www.pcper.com/news/Storage/WD-Re ... pport-TLER

Further more, is it necessary to have TLER (or similar) in a QNAP NAS to avoid the disk being dropped upon a read error?

[schumaku]

C'mon TLER is the "feature" which was available on some earlier WD desktop drives to time limit resp. almost disable the so-called deep-error-recovery time, using different ("normal") error recovery leading to RAID issues. The TLER code was removed on newer WD drives, so enabling the TLER-bit would not change anything - other than potentially causing new issues.

Every industry standard disk _does_ error recovery, ie. it will try to re-read a failed sector using different manufaturer-specific tricks for example. A sector causing problems several times will be dynamically substituted by one from the spare sectors reserved for this purpose.

So TLER is not a feature - it just "normalizes" the error recovery.

As these drives are dedicated for NAS RAID usage, the existance of an exhaustive deep-error-recovery is mosst unlikley. Theefore, there is no need to change the default HDD controller settings "enaling" the TLER bit...

WDC invented the deep-error-recovery on thier desktop drives - to aparently enahnce the HDD reliability. They used some more tricks ie. reading the data, and enhancing the sector - simply to avoid a sector replacement for more time. Over time, every sector recovered will not lead to a replacement, so the spare sector pool will not or much less used - under the line, this should cause less warranty replacements, too.

Conclude: No deep-error-recovey == no need for TLER.

To much marketing terms are creating confusion: WDC also used "Advanced Format" term for diks using 4k sectors for example.

[Briain]

Hi

I thought TLER (or equivalent; other manufacturers used to call it by other names) was enabled on all enterprise disks and that it was generally set to about 7 seconds. My understanding being that traditional hardware RAID controllers were pretty fiercely set and tended to chuck disks from the set if they were unreachable for about 10 seconds (that's why 7 was chosen for TLER). I did once spend a little while looking at the mdadm specifications (for those unfamiliar with it, mdadm is a Linux software RAID controller) but I couldn't see any obvious mention of any timers, so I don't know if (or by what process) a NAS would chuck a disk even if it did spend ages trying to re-read a duff sector (I've always meant to follow that research up and try to find out the full story) but I still always assumed that it was a desirable feature to have, as it would save a disk going nuts for ages trying to re-read when it just doesn't have to do so in a RAID set.

Bri

PS More interesting to me is that these new Red disks claim a non-recoverable error stats at <1 in 10 to the 14 (whereas the enterprise disks are much better at <1 in 10 to the 15); I'd love to know which technical aspects of the design or manufacturing process result in that difference between disks.

[foxbox]

Conclude: No deep-error-recovey == no need for TLER.

Thank you for your informative answer.
I was not aware that these disks did not do deep error recovery, one can only makes assumptions based on WD's marketing material.. But maybe they are afraid of consumers questioning why they have to pay more for "less" features if they spell out that NASware partly is about omitting 'desktop drive' features

PS More interesting to me is that these new Red disks claim a non-recoverable error stats at <1 in 10 to the 14 (whereas the enterprise disks are much better at <1 in 10 to the 15).

I've also been thinking about this since there are people claiming that the probability of a non-recoverable error during a raid rebuild are rising to troublesome levels with such data dense drives as 2-3 TB and above (with a non-recoverable error rate of <1 in 10^14) . And if a non-recoverable error do occur whilst rebuilding your RAID5 array and you already have a failed drive everything will go to **, or so some people say.

Please bear in mind that I have no professional understanding of this issue, but isn't this depending on how your RAID-controller operates and handles the error? Isn't it possible to skip over the 'error' and continue or have I misunderstood something fundamental about how RAID5 works?

[Briain]

Hi

Well I'm a microwave radio person and I'm most certainly not a computer scientist, so bear in mind that what I say could be incorrect, but my understanding is that it is the disk firmware that controls the re-reading process (in other words, it's the disk firmware that has to try and re-try to recover the data) and thus there's nothing the raid controller (mdadm in the NAS) can do to stop it doing that. My understanding - which again could be wrong - is that all desktop class disks will try and try (very hard) to recover data, since the alternative is just to lose it (i.e. for the computer to crash). As I say, my understanding is that TLER is a firmware feature to time that process out in about 7s and that it was introduced to prevent disks as being assumed faulty by hardware RAID controllers as found in proper servers. The situation with mdadm is less clear (well, less clear to me as I've never seen a definitive answer about all this) in that I don't know when (or how) a disk is regarded as defective, but all that said, I do believe that TLER is still a desirable feature; even if the disk is not at as much risk of being chucked out (as it would be in a hardware RAID controller), it would still prevent a NAS disk trying for too long when it can just mark the block as bad and retrieve the data from elsewhere in the set.

As I say, bear in mind that this is not my field of expertise and I would be very happy for anyone to come on and correct any erroneous assumptions in the above thoughts; I'd be grateful for any further information to clarify the processes in my own head (it has TLER set to about 0.1 picoseconds; if I can't figure something out within 0.1 ps, I instead think instead about something I maybe can figure out; normally beer)!

Bri

[P3R]

I've also been thinking about this since there are people claiming that the probability of a non-recoverable error during a raid rebuild are rising to troublesome levels with such data dense drives as 2-3 TB and above (with a non-recoverable error rate of <1 in 10^14) .

I think the URE-myth is based on this 2007 blog post that claimed RAID 5 should stop working (with desktop disks) in 2009 (when 2 TB disks was expected to be introduced). The many RAID 5-volumes running with 2 TB desktop disks (and 3 TB disks since late 2010) that haven't stopped working proved the prediction to be false.

The prediction was based on the assumption that every disk would have a URE at exactly 1 in 10^14 bits (12,5 TB) read. But that's not what the disk manufacturers specify. They specify that desktop disks will be able to read at least 10^14 bits before having a URE. Desktop disks are obviously generally better than the specification in this regard, since RAID 5 haven't stopped working yet.

Isn't it possible to skip over the 'error' and continue...

Not when rebuilding a degraded RAID-volume. At that point there is no data redundancy so all available data is necessary to correctly recreate the failed disk.

When a URE does actually happen during a rebuild it is a major problem and of course from time to time it does happen to some (but no way near as often as the myth claims). This is probably one explanation to why not all RAID-rebuilds are successful. This is yet another reason to why I promote backups in my signature below.

I don't think anybody outside the disk industry know if enterprise disks are really 10 times better than a desktop disk in this regard but that specification is part of why we pay more for enterprise disks. It's interesting to notice that Samsung did specify their desktop disks at 1 in 10^15 (incidentally they didn't offer enterprise disks). I've failed to find any official specifications on the low cost Seagate ST2000DL004 (that is a rebranded Samsung HD204UI) but it would be very interesting to see if they keep the higher specification or bring it down to protect their enterprise disk sales.

• For a business where data availability is more important than price, enterprise disks is the natural choice.
• For a home user on a tight budget desktop disk is most likely the way to go.

We all have to figure out where, in between those extremes, we fit in.

I use enterprise disks on my main NAS because I could get at them at a very reasonable prize but normally I'm a desktop guy. I am however very careful with my backups.

[Briain]

Hi

It is all very interesting in that the truth about the exact technical differences between enterprise and desktop disks seems to be a very closely guarded corporate secret. I've always wondered if they just select the best ones (perhaps based on lowest vibration and other other manufacturing tests) and install the enterprise version of the firmware on the best disks. It would be really interesting if the manufacturers could tell us the actual differences between models as without that information, we are all left just guessing (and some would argue that there is no difference other than a few FW tweaks to introduce TLER and maybe change seek performance characteristics); it would be nice to know the full story.

Curiously, Samsung did used to make enterprise disks (I have them running in my TS-659). Before I was a Qnap owner, I used to use ReadyNAS and always had the Samsung desktop disk (HD103UJ). They were very good and thus I recommended them to folks I know. As a result, there are a lot of folks I know who have been using HD103UJ in ReadyNAS and D-Link NAS boxes for years; none have yet failed. I did have an issue in a ReadyNAS NV+ where rebuilding the Twonky 5.0 database (that's the one where it generated over 3 GB of meta files for my 30,000 FLAC files; thankfully they reverted to using a SQLite db in Twonky 5.1 and above) used to result in a disk being chucked from the set. Curiously, I fitted enterprise disks and that stopped it happening; most odd.

Anyhow, I've been using the enterprise version of these disks (HE103UJ) in RAID 6 with my TS-659 for several years and they seem excellent (cool, quick and very quiet). They physically look identical to the HD103UJ disks, but when I looked at the spec sheets, the weights differed by a few grams (from memory, about 10g), so I wonder if they have higher specified motors to drive the heads. I guess that could make sense as in an enterprise situation, they'll be continuously seeking and thus be working a lot harder. Of course, my own suspicion is that that 10g anomaly might just have been a typo in one of the spec sheets, or maybe they just added a pointlss extra countersunk screw somewhere to make them 10g heavier than the desktop ones.

Bri

[P3R]

It would be really interesting if the manufacturers could tell us the actual differences between models as without that information, we are all left just guessing (and some would argue that there is no difference other than a few FW tweaks to introduce TLER and maybe change seek performance characteristics)...

There are almost always differences in specified power usage. Of course it could potentially be fake numbers but I don't buy that. It's too much of a conspiracy theory for my taste

I'm convinced there are actually mechanical differences in most, if not all, enterprise disk that aim to make them more reliable. On the other hand I don't think the price premium paid for an enterprise disk correspond to a similar increase in longetivity. I think desktop disks give more value for the money in non-critical environments.

Curiously, Samsung did used to make enterprise disks (I have them running in my TS-659).

Okay, I only looked at 2 TB and larger disks and couldn't find any enterprise disk from Samsung.

When I now do lock at the HE103UJ, the price premium for it seems to have been very small compared to the desktop version so maybe in that specific case they may actually have been very similar.

[TonyPh12345]

No, I'm not going on a tangent.

I bought a Toro lawnmower 10 years ago. It cost $799.
On the display floor right next to it was the commercial version for $1099.

When I compared the specs, I found there were *ZERO* differences between the two. Same engine, same fluid capacities, same deck construction & materials-- even the PARTS grids were identical (as evidenced by the Toro Document number.) Even the warranties' duration were the same. Of course, the labeling and badging were different, but that's irrelevant.

So what WAS the difference? The lower cost warranty term was cut in half if the unit was used in a commercial capacity.

Toro priced the commercial higher to cover their additional risk when the same unit was used in a commercial capacity and was far more likely to break down than a home-use unit.

I think the differences in "Enterprise" and "Desktop" CAN boil down to the same parallel, to a degree...

There may NOT be significant technical differences between the two: The manufacturer is just widening their risk profile.

[Briain]

Hi

You are possibly correct, but there are some tangible differences. Power usage is an interesting one in that I guess certain minor aspects of that could be set by the firmware (how aggressively you chuck the read heads about) but you'd think that the main motor current would be pretty consistent, so any changes in that aspect might imply a different platter motor (higher power and more possibly robust). That could explain the 10g difference between the desktop and enterprise Samsung disks. I guess there could also be closer tolerance bearings used in enterprise models, but again, that's highly speculative. Either way, it would be really nice to know some of these things as they are both interesting, and knowledge of them might help take some of the pain away when entering your card details to buy them.

Anyhow, talking of enterprise disks, I am wondering (just out of curiosity) if anyone has any opinions on the best of the two approved 3T enterprise disks for use in a Qnap (Seagate or Hitachi)? I was very tempted by the Hitachi 5K3000 (as mentioned in my post here) but they're about as common as rocking horse droppings and they have not yet been tested in a Qnap. So I guess that leaves the safe choices as Hitachi HUA723030ALA640 or Seagate ST33000650NS; anyone have any thoughts on which is best?

Incidentally, what interests me about the 5K3000 disks is the below information:

Power-on drive ready time (typical, sec) 20
Read/write (W) 6.2
Unload idle (W) 5.8
Low RPM idle (W) 3.2
Standby / Sleep (W) 0.9

I find hibernation to be a bit of a lost cause in my setup (they spin up and down all the time) so something which throttles back RMP (the above 3.2W figure) is quite an interesting idea; if that worked, I'd just disable hibernation.

Bri

PS I can't quite remember what the price difference was between HD and HE Samsung disks, but I think it was maybe something like £70 vs £120 (the latter being after a considerable search).

[paganizer]

I cant see the new WD red disk on the compatibility list for QNAP TS-EC879U-RP. Anyone know why?

Anyone tried the new disks with the QNAP TS-EC879U-RP (or similar NAS)? I'm a bit uncertain whether I should risk buying them or use hitachi or seagate enterprise disks

[schumaku]

I cant see the new WD red disk on the compatibility list for QNAP TS-EC879U-RP.

Because of it was not tested. WD does see the WD Red to be usd in small RAID with up to five HDD only:

WD Red

The new color of NAS.
WD Red hard drives are designed and tested for compatibility in the unique 24x7 operating environment and demanding system requirements of home and small office NAS.

WD Red NAS hard drives have been extensively tested for compatibility in 1-5 bay NAS systems in our FIT labs and AVL qualified with these key partners.

[P3R]

The WD Red disks are amazingly quiet. I have so far failed to hear them (2*WD30EFRX in RAID 1) at all over the fan noise in a TS-410.

[amnm]

I am happy about the release of the Western Digital RED drives. My question is what are the benefits of the the RED drives over the Seagate ones? I know that the RED drives are designed specifically for NAS, but I don't understand what is the benefit of that. The 3TB RED is still about $40 more expensive than the Seagate. Is there a justification for the price difference?

[skimmer333]

I am also curious as to which drive to purchase(either the WD30EFRX or the ST3000DM001)
It would be good to see a comparison/breakdown between the respective drives. (though im leaning towards the WD due to the cirping issue with the seagate (had returned 4 drives due to this, and they where on the latest firmware as well).