TLER / ERC / CCTL and RAID

[Tarrant1701]

Although QNAP's official hard disk drive compatibility list contains some consumer desktop drives, these drives often do not feature TLER or their equivalent feature (ERC / CCTL). My question is, are you using "enterprise" drives that feature TLER or regular desktop drives that do not have TLER?

I'm curious to see if users are reassured by QNAP's compatibility list, or if they use TLER-enabled drives to prevent RAID drop-outs. Thanks!

[Briain]

Hi

It's a great idea to highlight this. I think that Qnap should really make this much clearer for folks by splitting that table into two separate tables; one showing the enterprise drives and another for desktop ones.

I always strongly recommend people use the correct (enterprise) ones in RAID sets specifically because the TLER is set to suit that environment. Qnap should thus really split the tables and ideally also put a brief note explaining why there are two types of disk (and why folks should really consider buying the enterprise ones) as many folks seem totally unaware of the differences.

PS For anyone wondering what this is all about; see here for a reasonable description of TLER and here for my last post on that subject.

Bri

[Tarrant1701]

Thanks Bri!

As you pointed out in your other post, there are other differences between enterprise and desktop drives besides TLER. The unrecoverable error rate is one worth mentioning. Typical desktop drives have an UER of less than 1 in 10^14 bits. Compare that to enterprise drives designed for RAID and have TLER. Their UER is less than 1 in 10^15 bits.

This may not seem like a big deal, but consider this: a 2TB drive contains 1.8 x 10^13 bits. A QNAP TS-509 Pro with 5 populated bays would have 10TB or ~9 x 10^13 bits. In other words, 5 x 2TB drives would almost guarantee 1 UNRECOVERABLE error if you were using desktop drives with a UER of 1 x 10^14 bits! If one drive failed in a RAID5 array, you could say that statistically, you've just lost your data for good because you would hit 1 unrecoverable error while rebuilding.

[Briain]

Hi

I hope the top line (enterprise or RAID class disks) shoots up and the bottom line remains very low. If not, it shows the need for Qnap (and other NAS vendors) to explain the difference right below their recommended disk tables!

Yes, one of my friends worked that out whilst we were consuming fine ale one evening. As he pointed out, hard drive sizes have exploded over the very recent past but that UER hadn't reduced in proportion to disk size increasing and thus it's the statistical likelihood that needs to be considered. He thus postured that it's now quite a lot more likely in the life of a something as big as a 1T drive and this helped prompt my decision to now use a ReadyNAS Duo (in RAID 1) as my backup device for key data on my Qnap. I think it's a great shame that the NAS manufacturers don't explain this (and TLER) to folks along side the hardware compatibility list table as it's very important. I believe that folks should chose these enterprise (RAID class) disks unless they are operating on a shoe-string budget. For SOHO and above business users, is't a 'no brainer' and it concerns me that the less technical business user will likely buy desktop disks; there is simply nothing to say why they shouldn't and that really should be addressed pretty swiftly!

Incidentally, that same friend had a RAID 5 disk failure then slotted a new one in. The stress of rebuilding the set prompted the 'death' or read errors from another old disk (can't remember which) and that was the end of the RAID set! I have heard of others having the exact same problem and for that reason alone, I'm now using my Qnap with 4 disks set to RAID 6 (with the correct enterprise disks, of course) and my RAID 1 backup (again with the correct enterprise disks) will hopefully not ever be required unless I get a major catastrophe with the RAID 6 Qnap (like the shelf falling off the wall and breaking all the disks, or, of course, fire/theft).

The off-site USB backup is my third line of defense; it is stored in a fire-proof box in my parent's cellar.

My end plan is to locate the ReadyNAS in my parents house and rsync the Qnap to it over the www.

Paranoid; me; naw, surely not!

[ginfonic]

Hi
If you don't mind I make my contribution to this vital issue.
TLER... Of course it is important feature of enterprise drives.
But... could you tell me what gave you the idea of TLER support in QNAP NASes?
Yes, support of TLER is an everpresent feature of enterprise HARDWARE RAID controllers. But QNAP have only SOFTWARE Linux-based RAID. Does it support TLER or doesn't, who knows.
There was discussion in other thread where this question was asked. QNAPJason answered: "...QNAP NAS use software RAID, which monitors the RAID error recoveries based on the detections..." What does it mean "detections" and whether TLER supported he didn't answer.
Googling a little I found different opinions but more of them says that there is NO TLER support in Linux-based RAID arrays. There is one of them, for instance.
In other words we have nothing more (and nothing less) but little black Linux-powered PC with 2-4-6 HDs attached. What HDs should we use in this case? I seem desktop ones is the answer.

[Tarrant1701]

Good question regarding whether QNAP actually supports TLER. But remember the other advantage of "enterprise" grade drives and lower rates of unrecoverable error rates. That still is a consideration.

[ginfonic]

By the way up-to-date WD desktop HDs have the same UER as enterprise ones -- < 1 in 10^15 bits. Thus it's not an unique feature of enterprise HD.

[Tarrant1701]

That is a half-truth. While the UER is < 1 in 10^15 for Caviar Green drives from WDC, it is still 10^14 for:

1. Caviar Blue:
http://www.wdc.com/wdproducts/library/S ... 701277.pdf

2. Caviar Black:
http://www.wdc.com/wdproducts/library/S ... 701276.pdf

3. Seagate Barracuda XT
http://www.seagate.com/docs/pdf/datashe ... uda_xt.pdf

4. Seagate Barracuda 7200.12
http://www.seagate.com/docs/pdf/datashe ... 200_12.pdf

I would certainly call these "up-to-date" desktop HDs, and the number of desktop drives that have UER of 10^14 exceeds the number that have 10^15.

[Briain]

Hi

Actually, I was having a ponder on this and TLER is set on the disk firmware itself. I'm familiar with engineering and electronics, but not so much about the computing side of things, but it stands to reason that as TLER times are set in the disk firmware, the disk firmware itself will surely determine how long it tries to read any corruption (otherwise you'd not set it in the disk firmware, you'd instead set it in the BIOS or OS). The question is really what actually happens when a problem exceeding the set TLER time does occur within a disk; what is the next stage in correct process when that actually happens? It would seem logical to assume the firmware sends a disk fail signal condition back to the BIOS, and if that's true, the question would then be whether that condition can be ignored by the BIOS, and instead of using that standard fail condition to determine a disk is in trouble, the software RAID controller element within the OS would then have to include an alternative mechanism (and timer) to decide when a disk has failed. If that were the case, it wouldn't matter which disk you used as a virtual TLER would be set by Qnap. To me, that sounds a little unlikely and thus I'd assume that TLER does matter.

When I get some time, I'll go and find out how the disk fail process is normally meant to work and report back when I do. Knowing what's meant to happen will help us understand whether the condition is easy to ignore or not. At the very least, it'll certainly give us a far more exacting technical question to then ask the Qnap folks

Bri

PS Does anyone reading this happen to know the 'industry standard' handshaking process for TLER based disk rejection; is it something really basic like an earth condition that generates an alarm via BIOS? If so, please chip in as it'll save me ploughing through tonnes of technical documentation.

[cstork]

Hi Bri,

Thanks for looking into this. This thread in the Seagate forums

http://forums.seagate.com/t5/Barracuda- ... /m-p/54082

should contain some interesting pointers for this discussion. I guess the ATA specs should be relevant here.

(It's nice to see that Seagate allows to dis/enable ERC according to standard whereas WD is behaving very consumer unfriendly. I'm intending to buy Barracuda LP HDDs, but I'm still not sure if in the end it's better to enable or disable ERC with Qnap's software-based RAID...)

Chris

[Moogle Stiltzkin]

Read this article where they question whether tler was needed for NAS or not
http://www.smallnetbuilder.com/nas/nas- ... r-raid-nas


From my personal experience having owned western digital enterprise raid drives, and a Samsung desktop F3 HD203WI drives which i both use for raid in my TS-509 Pro and TS-659 Pro II respectively, i can tell you that the desktop drive works just as well as .... at cheaper a price to boot.

[lucas72]

Also read here and here.

[Briain]

Hi

Quite an old thread! Since my initial posts, I did a lot more research and realised that it's still a better option to have TLER enabled on any disks that are in a RAID set. It's really a feature designed to prevent long retries causing a hardware RAID controller to chuck a disk out of a set (whereas most NAS's use mdadm, which is a software RAID controller), but even though the NAS will not chuck a disk after a few seconds, the TLER is still useful in that it will prevent a disk suffering an unrecoverable read error from spending ages going catatonic trying and trying to recover data, when instead it can just give up (and the data will be recovered from the another disk, then the area marked bad and the data written elsewhere on the disk).

So there's maybe no need to use enterprise disks in a NAS, but the TLER feature is desirable and tends to only be available on enterprise disks (so they are better, but they are also very pricy). What I have never been able to find out (on the manufacturers' sites) is whether that is the only difference between desktop and enterprise disks. It shouldn't be, as it most certainly doesn't justify the huge price difference, so I wonder if they measure another parameter to select them? One I could think of would be vibration levels; they could use the lower vibration ones for enterprise disks and the poorer ones get used for desktop disks. That would seem a plausible criteria that you'd think could be relatively easily adopted during the manufacturing and testing process (and would relate to the longer life expectancy for enterprise disks) but it is just a wild guess!!

Bri