Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Unfortunately I didn't examine the issue any further, and I don't even use the card for internal drives any more.

But if others have this kind of setup running then there shouldn't be a problem.
 
Unfortunately I didn't examine the issue any further, and I don't even use the card for internal drives any more.

But if others have this kind of setup running then there shouldn't be a problem.
What exact card did you use?

I'm wondering if it's the firmware, or possibly they didn't use the reference design (i.e. trying to reduce the part count to save costs - there's not much to them though). :confused:

There's other cards, but they're all eSATA ONLY, and means an external enclosure would be needed.
 
Hello,

Using a PCI-e card like the sil3132, are there significant speed differences between an eSATA external drive, or that same drive plugged in one of the MP's internal bays?

Loa

P.S. I can't test it myself as I only have RAIDed drives inside my Mac Pro.
 
What exact card did you use?

I'm wondering if it's the firmware, or possibly they didn't use the reference design (i.e. trying to reduce the part count to save costs - there's not much to them though). :confused:

It's this card: Digitus DS-30102

Looks pretty much like the reference design, but who knows? :rolleyes:
Does only work with Sonnet drivers, even the drivers they offer don't work with OS 10.6.
 
what about SoftRAID?

In this entire thread I haven't seen anyone mention the "other" (i.e. not Apple) software RAID.

Is it because nobody wants to pay extra for software when Apple's is "good enough"? Or is it because everyone thinks hardware RAID is the way to go?

Yes I know there are limitations. E.g. no RAID-5. But Apple's software doesn't support that either.
 
Is it because nobody wants to pay extra for software when Apple's is "good enough"? Or is it because everyone thinks hardware RAID is the way to go?

I haven't heard about this before, but to be honest, what is the major advantage of SoftRaid that makes me wanna pay $70 over using a software RAID that is already implemented, hence fully supported and widely tested in the operating system?
 
It probably wasn't mentioned because I was only wondering about RAID 0 which is supported by Apple.
 
SoftRAID is $129. It claims: Flexibility, Reliability, and Usability are the three main distinctions between Apple's Disk Utility and SoftRAID 3.

I haven't used Apple's RAID nor SoftRAID. I was hoping someone who has used both would chime in and report if it's "worth it". Obviously it can't be too popular or there would have already been a plethora of postings extolling its virtues.
 
Using a PCI-e card like the sil3132, are there significant speed differences between an eSATA external drive, or that same drive plugged in one of the MP's internal bays?
For the same card, No. But you will get into issues with the PCIe lane count (i.e. 1x lanes, and PCIe spec. 1.0 = 250MB/s per lane). But it's still 250MB/s, which is the limit of PM chips, and more than enough for mechanical drives.

SSD OTOH, is worth finding a SATA 3.0 model (6.0Gb/s), and it will use additional lanes as well, and likely be capable of running PCIe spec. 2.0 (500MB/s per lane). Much better for SSD use. There's just no internal port versions AFAIK that will work under OS X.

There are other cards with more lanes (i.e. 4x port models, but are substantially more expensive, but uses 4x lanes - still SATA II = 3.0Gb/s spec.).

It's this card: Digitus DS-30102

Looks pretty much like the reference design, but who knows? :rolleyes:
Does only work with Sonnet drivers, even the drivers they offer don't work with OS 10.6.
I can't tell. Those chips are on so many different brands, makes it more difficult as well. And without the card, I've no way to even attempt to find out who actually manufactured the thing (hopefully, it has a manufacturer code). :confused:

In this entire thread I haven't seen anyone mention the "other" (i.e. not Apple) software RAID.

Is it because nobody wants to pay extra for software when Apple's is "good enough"? Or is it because everyone thinks hardware RAID is the way to go?

Yes I know there are limitations. E.g. no RAID-5. But Apple's software doesn't support that either.
I've not used it, but if it does the same thing that OS X's Disk Utility does, why spend the additional money?

I'd skip it, and go for a hardware solution (which is what I do, as software isn't actually capable of operating parity based arrays, if it has the capability or not - write hole issue is left unsolved).

I haven't used Apple's RAID nor SoftRAID. I was hoping someone who has used both would chime in and report if it's "worth it". Obviously it can't be too popular or there would have already been a plethora of postings extolling its virtues.
I'm not seeing anything that's actually different (allows it to claim it's more stable, just a claim).

But I've not used it, so I can't be 100% sure, but my instincts are telling me it's a false statement these days. IIRC, someone had mentioned some time ago, that it was true under earlier systems (PPC days I think), but that's no longer the case if I'm remembering the post correctly (it's here in MR somewhere, but I can't recall where and am too lazy to search for it right now :eek:). ;) :p
 
someone mentioned that for raid0 in osx, that you need to create the array with disk utility before you install the osx.

in my case, i am using raid0 for scratch disk. so i am NOT running raid0 on my OSX OS, but instead using it as internal HDs for expandable drives.

is it possible to non-destructively make raid0 in osx without doing any harm to osx? or will i need to reformat my osx as well?

also, if i want to expand say a 2TB raid0 to a 3TB raid0 by adding an additional 1TB drive, can i just "add" it in disk utility or will i need to back up the 2TB to another drive, wipe out the raid0 to a raid0 with 3TB, finally moving data back over to the new raid 0?
 
also, if i want to expand say a 2TB raid0 to a 3TB raid0 by adding an additional 1TB drive, can i just "add" it in disk utility or will i need to back up the 2TB to another drive, wipe out the raid0 to a raid0 with 3TB, finally moving data back over to the new raid 0?

I'm pretty sure you're gonna have to back up the 2tb array, wipe it, create the 3tb array, then move data back.
 
someone mentioned that for raid0 in osx, that you need to create the array with disk utility before you install the osx.

in my case, i am using raid0 for scratch disk. so i am NOT running raid0 on my OSX OS, but instead using it as internal HDs for expandable drives.

is it possible to non-destructively make raid0 in osx without doing any harm to osx? or will i need to reformat my osx as well?
If you leave OS X's disk alone (stripe set is separate from the OS disk), then the OS disk will be fine. No data loss will occur.

The drives used for the stripe set however, WILL BE WIPED as a result of the initialization process.

also, if i want to expand say a 2TB raid0 to a 3TB raid0 by adding an additional 1TB drive, can i just "add" it in disk utility or will i need to back up the 2TB to another drive, wipe out the raid0 to a raid0 with 3TB, finally moving data back over to the new raid 0?
Yes, you'd have to make a full backup, add the disk and create the array (no way to do online expansion), and the data is wiped during the initialization phase (just as it is with the initial setup).

Online expansion is possible with proper RAID cards, but not usually with software based arrays (there are a few proprietary products that claim they can, such as some NAS products that are software based). But I've never seen it with OS based software RAID.
 
ahh i see,.

what if i have a software raid, ie raid0 that is seperate from OSX partition

and my osx gets formatted, will i lose the raid partition?
in other words if my osx non-raid HD gets corrupted, will the raid0 go down with it?
 
ahh i see,.

what if i have a software raid, ie raid0 that is seperate from OSX partition

and my osx gets formatted, will i lose the raid partition?
in other words if my osx non-raid HD gets corrupted, will the raid0 go down with it?
No.

If the OS X disk goes, the RAID will remain in tact. The reverse is the same as well; array gets wiped, the OS X disk will still be in tact, and working.

This is the biggest advantage of running a separate OS disk from the array, and why it's done. :D

A hardware implementation would be considerably more robust (i.e. can move it from system to system, capable of increased throughputs, and capable of levels not possible under OS X's Disk Utility, can be accessed via multiple OS's), but also have to pay for it. :rolleyes: :p
 
...you typically have to run enterprise HDD's rather than consumer models (which can be used in a software RAID, as the control is retained by the system rather than the card). So the recovery timings in the consumer models will work.

That said, enterprise drives are a good idea anyway, as they're rated better, and have additional sensors that prevent the heads from physically crashing into the platters.

nanofrog, this statement of yours from April prompts a related question:
Can it ever make things worse to use enterprise HDDs in RAID 0?​
If the enterprise HDD quits error recovery at 7 seconds and sends a message to the (hard or soft) RAID controller to work around the error, then since the RAID 0 striping gives the controller no such ability what will it do? Will the controller message the HDD to resume recovery and will the HDD obey? What if the controller doesn't understand the (TLER) message from the HDD or ignores it? Do you recommend disabling TLER on enterprise HDDs in RAID 0? Can it always be disabled?
 
Can it ever make things worse to use enterprise HDDs in RAID 0?​
I've never seen that particular case (error a result of enterprise disks, and consumer units would have been fine). Usually it's a disk failure, and the consumer models are more likely to go within a period of time. Plenty of proof of that, not just my observations.

The biggest issue to me, is there's no parity data to rebuild a bad sector from with a stripe set, but that's the compromise of that particular level (issue exists in either a software or hardware implementation). It's not the drives' fault, but the level selected by the user.

Personally, I usually skip over stripe sets as I need redundancy, though they do have their place. Basically low budget performance (redundancy either not required, or insufficient funds to obtain it), but can be effective if the bare minumum precaution is taken (backup system, and it's kept up with). In such a case, the compromise for that low cost is time necessary to rebuild the array when it fails.

If the enterprise HDD quits error recovery at 7 seconds and sends a message to the (hard or soft) RAID controller to work around the error, then since the RAID 0 striping gives the controller no such ability what will it do? Will the controller message the HDD to resume recovery and will the HDD obey? What if the controller doesn't understand the (TLER) message from the HDD or ignores it? Do you recommend disabling TLER on enterprise HDDs in RAID 0? Can it always be disabled?
The TLER values has to do with the disk's ability to remap a bad sector.

In the case of a single disk (0,0 settings), the disk will become unavailable if an error occurs, as it's trying to remap that bad sector. As a result, the system is temporarily unresponsive. In instances where the remap is unsuccessful (i.e. many bad sectors), you're libel to get an error message from the OS after some period of time, as the disk has yet to "recover", and become available to the system.

This is necessary, as there's no parity data in single disk operation.

In the case of a RAID set, if this method of recovery were allowed to happen, the set would also become unusable for some unknown period of time. It's different in this case, as it's expected that the array is available to multiple users and/or needs to be available as much as possible (redundancy). Think in terms of IOPS for usages like database systems.

To combat this, the enterprise disks have different settings, which allows the recovery operation time out at a set period of time (7 seconds during writes), to keep that array available. This isn't a problem as most RAID levels either use parity data, or another disk the data's duplicated on the corrupt data can be recovered from (rebuilt automatically). This is why most RAID card's won't even work with consumer disks (unstable at best).

A stripe set (RAID 0) is the bastard child of RAID, as it has no redundancy whatsoever, and why it's not used for high availability scenarios. So if you're running this level, it's expected the user realizes the limitations. You still have some ability to recover, but not endless (7 second limit, unless you adjust the values on the drives). If it goes past that time limit, the data will be corrupted. Basically, user peril/"learn the hard way" if they don't understand this limitation.

Hope this helps. :)

BTW, wiki has a page on TLER if you want to research this in further depth. ;)
 
...A stripe set (RAID 0) is the bastard child of RAID, as it has no redundancy whatsoever, and why it's not used for high availability scenarios. So if you're running this level, it's expected the user realizes the limitations. You still have some ability to recover, but not endless (7 second limit, unless you adjust the values on the drives). If it goes past that time limit, the data will be corrupted. Basically, user peril/"learn the hard way" if they don't understand this limitation.

Hope this helps. :)

BTW, wiki has a page on TLER if you want to research this in further depth. ;)

Thanks nanofrog for jumping back into the strand. Actually your statement quoted above does imply that the enterprise class hard drive with its 7 second limited error correction will, in terms of error correction, be worse than the desktop class hard drive for RAID 0.

I had studied the Wikipedia page on TLER and also a Samsung page on their CCLR but don't understand the interactions of hard drive and controller, leading to my original question.

Yes RAID 0 is a bastard, and I'm afraid statements about RAID simpliciter sometimes overlook RAID 0 and become false. A writer of the Wiki page is obviously pissed that the utility WDTLER doesn't work on the newest WD drives, so TLER can't be disabled on their new enterprise drives. He/she writes:

"...having the choice of changing the TLER option gives the user the flexibility of purchasing the high performance drives that have TLER enabled and that are designed to run in a RAID array and using them as stand-alone single drives in their desktop computers and disabling the TLER option to enable better data recovery and lower chance of corruption and drive failures."

Here the writer carefully separates having "TLER enabled" from "designed to run in a RAID array" (durability which entails higher heat, noise, and cost) of the enterprise class drive. I think RAID 0 drives want TLER disabled for exactly the same reason that a stand-alone drive does. The writer goes on to slam WD with the conclusion:

"[New enterprise] RE disks are only suitable for RAID arrays and [new desktop] Caviar are only suitable for non-RAID use."

But here I think the first word "RAID" needs to be qualified "except RAID 0", and maybe the second word "RAID" too.

My purpose for joining drives in RAID 0 is for playback of some video which will stream at about 800 Mb/s (100 MB/s). A single drive can't read fast enough. Maybe two in RAID 0 can. Maybe three. If not three, I'll quit. Samsung 1 TB hard drives here cost 45% extra for enterprise class. That's fair, but if the 7 second error correction limit can't be disabled I'll go with desktop class. Backing up is fun to do.
 
Thanks nanofrog for jumping back into the strand. Actually your statement quoted above does imply that the enterprise class hard drive with its 7 second limited error correction will, in terms of error correction, be worse than the desktop class hard drive for RAID 0.
:cool: NP. :)

RAID is complicated, and can get confusing rather quickly. Available online resources can be a bit too simplistic, and lead someone into trouble IMO.

At any rate, I get what you're meaning about the factory timings in the enterprise models.

But there's two aspects of information that're being missed (one is my fault :eek:).
  1. You can change the timings on the enterprise disks (REx models).
  2. Even if you don't you won't see that issue like you would with consumer models (and why I forgot to mention #1 :eek:). It has to do with the fact those disks are meant for the abuse RAID hands off to the disks via the additional sensors (i.e. if it senses the vibration is getting too high, it will act to prevent the heads from impacting the platter causing physical damage = bad sectors result).

I had studied the Wikipedia page on TLER and also a Samsung page on their CCLR but don't understand the interactions of hard drive and controller, leading to my original question.
Understandable.

What's worse, the exact procedure/methods used internally in the disks are still somewhat proprietary (i.e. what sensor at what value will it temporarily disable the disk's operation in order to stabilize it). Yes, SATA is a specification (open standard), but what happens in the controller board's interaction with the platters, spindle motor, and stepper motor (operates the armature the heads are mounted to), can vary from drive maker to drive maker (though the basic process isn't that different between them). SATA has commands that can be sent bidirectionally (i.e. disk sends error message to the system). Think SMART data.

So long as a disk sends/receives data within the SATA IO specification, it's deemed compliant. What happens on the other side (disk controller), may fall under Intellectual Property = proprietary (Patents). Though there is a lot of commonality (basic requirements to operate the mechanics to get/transfer data), the way they do it may not be exactly the same. Same goes for platter production (density for example).

Yes RAID 0 is a bastard, and I'm afraid statements about RAID simpliciter sometimes overlook RAID 0 and become false. A writer of the Wiki page is obviously pissed that the utility WDTLER doesn't work on the newest WD drives, so TLER can't be disabled on their new enterprise drives. He/she writes:

"...having the choice of changing the TLER option gives the user the flexibility of purchasing the high performance drives that have TLER enabled and that are designed to run in a RAID array and using them as stand-alone single drives in their desktop computers and disabling the TLER option to enable better data recovery and lower chance of corruption and drive failures."

Here the writer carefully separates having "TLER enabled" from "designed to run in a RAID array" (durability which entails higher heat, noise, and cost) of the enterprise class drive. I think RAID 0 drives want TLER disabled for exactly the same reason that a stand-alone drive does. The writer goes on to slam WD with the conclusion:

"[New enterprise] RE disks are only suitable for RAID arrays and [new desktop] Caviar are only suitable for non-RAID use."

But here I think the first word "RAID" needs to be qualified "except RAID 0", and maybe the second word "RAID" too.
WD used to allow the TLER values to be adjusted in all of their disks (initially didn't have RE = RAID Edition disks for every niche, such as in the Green series). That's now changed, so only the RE series (RE4 to be a bit more specific) allow the TLER values to be adjusted by the user with the utility. The consumer models have had that ability removed. You can still do it with older drives, but no longer with anything new in this segment (standard Blue, Green or Black, or other specialty disks; anything not identified specifically REx in the model).

This has pissed some people off, but common sense should have led them to realize this "freebie" wouldn't last forever. They're after profits, and allowing this to continue on the consumer lines cut into that, as users weren't buying the RE models.

Some got burnt too, depending on what they were doing (i.e. using consumer models for a primary array).

Consumer GP disks were commonly used as primary (on-site) backups once the TLER values were adjusted, as they were suitable for that, and corporations are noticing power bills and HVAC requirements (especially during an upgrade to the existing facilities, new facilities, or a break down of existing equipment - it's expensive).

My purpose for joining drives in RAID 0 is for playback of some video which will stream at about 800 Mb/s (100 MB/s). A single drive can't read fast enough. Maybe two in RAID 0 can. Maybe three. If not three, I'll quit. Samsung 1 TB hard drives here cost 45% extra for enterprise class. That's fair, but if the 7 second error correction limit can't be disabled I'll go with desktop class. Backing up is fun to do.
2x disks will suffice for this, especially if the individual performance is at/over 100MB/s each (allows you to go past the 50% capacity mark). Think 2TB models that run at 7200rpm (greens aren't that fast).

With slower disks, it will still work, but you have to be mindful of the capacity, as data transfers slow down once you hit the inner tracks on the platters (same rotational speed, but shorter tracks = less data read/written per rotation = reduced throughput). Additional drives help with this, but it's also at additional cost.

In your case, you probably don't need to bother with enterprise grade disks (assuming you're using the SATA ports on the logic board, not a true RAID card), especially with a proper backup in place.

Yeah, it's a PITA to rebuild an array by hand off of backups, but it's an acceptable compromise to keep the costs as low as possible.

But the enterprise disks would reduce the statistics of this happening, and the values can be adjusted with the TLER Utility (not just disabled, but actually changed to other values). I know I've seen people complain this isn't possible, but it is. You just have to enter the correct command (values are contained within the command). No values, means defaults will be used.

Hope all of this helps. :)
 
nanofrog, your information is very useful. If the error recovery timings of Enterprise HDDs can be changed, then my worry concerning RAID 0 disappears. The Wikipedia writer doesn't believe that the new WD Enterprise drives can have their TLER settings changed by the WDTLER utility. He/she wrote:

"Western Digital now claims that using the WDTLER.EXE tool on newer drives can damage the firmware and make the disk unusable. The WDTLER.EXE tool is no longer available from Western Digital, and new disks will not be able to have the TLER setting changed. RE disks are only suitable for RAID arrays and Caviar are only suitable for non-RAID use. The utility still works for older disks."

You contradict this when you write:

"That's now changed, so only the RE series (RE4 to be a bit more specific) allow the TLER values to be adjusted by the user with the utility."

The Wikipedia page (which you originally recommended) needs your input. Incidentally, in this one place you wrote that only RE4 series drives allow the timing to be set, but elsewhere you say REx. The 1 TB Western Digital drive now available here is the RE3 WD1002FBYS. Does WDTLER work with it?

And what about Samsung Enterprise class hard drives? (Today's price here for the SpinPoint F1 HE103UJ is 74% the RE3 WD1002FBYS price.) Can the error recovery timing of the SpinPoint F1 HE103UJ be set by the user?

(I won't ask about setability on Seagate Enterprise class hard drives because they're way too expensive here, but someone else should ask about this.)

If we agree that there are uses -- single drive; RAID 0 array -- for which the factory TLER settings on Enterprise class drives are bad, then we need to find out which Enterprise class drives offer TLER setability and how.

Thanks.
 
online purchase of enterprise disks?

All this discussion begs the question of where to buy "enterprise" disks. Even if a consumer oriented vendor like Newegg had the disks, they might well be preloaded with random firmware for various OEMs.

So, who can we trust online?
 
nanofrog, your information is very useful. If the error recovery timings of Enterprise HDDs can be changed, then my worry concerning RAID 0 disappears. The Wikipedia writer doesn't believe that the new WD Enterprise drives can have their TLER settings changed by the WDTLER utility. He/she wrote:

"Western Digital now claims that using the WDTLER.EXE tool on newer drives can damage the firmware and make the disk unusable. The WDTLER.EXE tool is no longer available from Western Digital, and new disks will not be able to have the TLER setting changed. RE disks are only suitable for RAID arrays and Caviar are only suitable for non-RAID use. The utility still works for older disks."

You contradict this when you write:

"That's now changed, so only the RE series (RE4 to be a bit more specific) allow the TLER values to be adjusted by the user with the utility."
The RE4 statement meant the REx drives, RE4 being the most recent so far. I can adjust those models, though as mentioned, if you're not familiar with using the WDTLER utility, you may think that it's only possible to Enable/Disable the setting, not adjust the actual settings.

I have RE3 series in my personal system, and can adjust them. The Velociraptor (300Gb) and Caviar Blacks (1TB) can as well, but they're a bit over a year old now (model released before the RE4-GP model).

The Wikipedia page (which you originally recommended) needs your input. Incidentally, in this one place you wrote that only RE4 series drives allow the timing to be set, but elsewhere you say REx. The 1 TB Western Digital drive now available here is the RE3 WD1002FBYS. Does WDTLER work with it?
Sorry about the confusion, so see above (REx through RE4, which is the current revision number, is what I was trying to get across to you).

And what about Samsung Enterprise class hard drives? (Today's price here for the SpinPoint F1 HE103UJ is 74% the RE3 WD1002FBYS price.) Can the error recovery timing of the SpinPoint F1 HE103UJ be set by the user?
I don't deal with Samsung disks, as their consumer disks suck. To me, that's an indication of the enterprise models as well. I have the same opinion of Hitachi. But the variance between model years is enough that one bad year or so may be followed with a really good one or two. Hard to say, but I then look at warranty periods, and WD's still offering 5yrs on their enterprise disks, which makes me less hesitant (easy to get a replacement in the US if something does go wrong, as I have had to do it not that long ago, and it was an enterprise disk, but the only one out of many in the last couple of years or so).

Nor have I ever seen or heard of such a tool for Samsung or Hitachi's products. If it exists, it's kept internally within the respective company.

(I won't ask about setability on Seagate Enterprise class hard drives because they're way too expensive here, but someone else should ask about this.)

If we agree that there are uses -- single drive; RAID 0 array -- for which the factory TLER settings on Enterprise class drives are bad, then we need to find out which Enterprise class drives offer TLER setability and how.
AFAIK, Western Digital is the only disk maker to offer such a utility. I know for certain Seagate doesn't do this (only works with the card maker), which means either a card or disk firmware upgrade is offered, if anything. It's the same for Samsung and Hitachi (internal, and they release a new firmware if anything, as the card maker is more likely to make an adjustment to their firmware or just state whatever P/N is incompatible with their product).

Even if a consumer oriented vendor like Newegg had the disks, they might well be preloaded with random firmware for various OEMs.
Look for WD, Seagate, .... OEM disks, not from sources like HP,... (system vendors), that may have a specific firmware revision that's meant to work with their offered card/s.

So far, I've not had any issues from newegg, or other vendors this way (provantage does offer HP for example), but also OEM disks with the drive makers' names on the disks. Most of the time, the vendor labeled disks aren't an issue either, particularly HP (usually SAS models, as you can find on provantage.com). But it is possible, and best to avoid them if you're uncertain.

Shouldn't be a problem if you pay attention to the details. ;) If you're not sure, ask. If I don't see it, someone is surely willing to help, as that's what a forum is for. :D

Hope this helps. :)
 
How reliable would a RAID 0 be in use for time machine?
Well, it's not used as often as a primary array which lowers it's frequency of failure a bit (assuming the spin-up doesn't go awry, as this array is usually spun down). Personally, I still wouldn't trust it. Especially if the primary is a stripe set as well (increases the chance that both arrays fail simultaneously = data is DOA).

JBOD is a better implementation IMO (can be done with no additional cost just by configuring existing disks), as worst case, you only loose the data on the failed disk, not all of it (assumes the primary source is also gone when a disk fails in the JBOD).

If your data is super critical, you'd want to consider other levels for both a primary and backup array. Which level however, would depend on specific details.
 
Contradictions fly around this topic. Anonymous (and undated) Wikipedia TLER writer says new WD Enterprise drives will not be able to have the TLER setting changed. A WD Knowledge Base page says the TLER feature on Western Digital RAID Edition SATA hard drives cannot be disabled. nanofrog says he can adjust the TLER setting on his WD RE4 drives.
[added 4 hrs later: The RE4 is an attractive drive with its 64 MB cache and low power consumption. Can the TLER firmware setting on currently produced RE4's be adjusted with WDTLER?] There's internet buzz over exactly when in 2010 WD made it impossible to use WDTLER to modify the firmware in their RE drives. This is a rumor forum too.

I don't think disabling TLER is fundamentally different from adjusting its times to 0, which codes infinite time or maybe 200,000 milliseconds if the drive itself has some such limit.

There's no agreement over whether TLER enabled (set to 7 sec) is bad for drives in RAID 0. I think it is bad. I think the RAID 0 error correction strategy should mimic the single drive's. But there's no agreement over the single drive either. The anonymous Wikipedia TLER writer says that for single drives "disabling the TLER option [enables] better data recovery and lower chance of corruption and drive failures." Another WD Knowledge Base page says that "a drive with TLER enabled will work with no performance decrease when used in non-RAID environments."

I don't deal with Samsung disks, as their consumer disks suck. To me, that's an indication of the enterprise models as well. I have the same opinion of Hitachi. But the variance between model years is enough that one bad year or so may be followed with a really good one or two. Hard to say, but I then look at warranty periods, and WD's still offering 5yrs on their enterprise disks, which makes me less hesitant (easy to get a replacement in the US if something does go wrong, as I have had to do it not that long ago, and it was an enterprise disk, but the only one out of many in the last couple of years or so).
For what it's worth, Samsung Enterprise drives (including the inexpensive HE103UJ) have 7 year warranty.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.