• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

TLER and other Esoterica:

Rowebo

Junior Member
Seeking answers on a couple of fronts:

1. Shopping for Drives, and have had good luck with WD lately. Was scoping the Caviar Black series, but ran into reviews in which -- first, you need to use a separate utility to turn on TLER to use Black Eds. in RAID; and second, WD has now rewritten firmware to prevent the change.

Questions:

1. Should I be annoyed with WD for this sort of behavior? The RE drives shouldn't cost much more, but often do in practice, as Newegg etc, push the blacks. I'm not going to RAID the drives in question initially, but hate the absence of versatility down the road. If WD is protecting folks from data loss, that's one thing, but if they're protecting their profit margin, well, there are others out there. And given the prevalence of RAID discussions, it seems likely that lots of black drives will get used in desktop arrays, so why hurt your sales by making all this trouble -- why not just give the masses TLER?

There seems to be a lot of shopping for drives that can still be manually TLER'd, which tends to suggest that I shouldn't be all that happy with WD. But it's a tad hard to tell.

2. That said, there are suggestions here . . .

http://forums.anandtech.com/showthread.php?t=2041288&highlight=tler

. . . that if you use a particular controller setup, you don't care, presumably because you're running an independent solution. The implication is that this is unique to the ICH10R ports, and that other on-board RAID solutions would still suffer, correct?

3. On the same thread, there's a brief discussion of short-stroking. In addition, I've recently become aware of drive alignment, and just tried aligning the boot drive on a new Win 7 install with Diskpar. Not sure how much of the snap I'm seeing is Win 7 vs XP (most test reports say not that much), how much is because of the 64 bit/4 gigs of memory benefit, and how much is the alleged 20-30% benefit of alignment, but overall, I'm pleased. And the only apparent downside is the loss of the ability to use partitioning utilities -- which seem not to recognize the aligned partition as NTFS.

All of which has focused my limited attention span on the idea that there may be I/O gains to be made short of dumping large sums into an SSD. Might there be a good tutorial out there regarding drive performance and tweaks so that I might get a bit more up to speed?

Thanks in advance for any widsom . . . off to driveway snow removal detail . . .
 
TLER is only good on raid systems because the raid controller needs to maintain the decision to handle bad sectors/remapping/dropping the drives.

conversely in jbod/1 drive mode - you really want that drive to try really hard to recover the data since it is not part of a raid.

Raid controllers say under any condition if i do not get a response from a drive in 8 seconds - the drive is bad - drop it.

Non-TLER drives will (especially when busy) take up to 30 seconds or more to keep trying.

TLER for raid use only.
consumer for non-raid.

And the RE3/RE4 western digitals are binned like everything else to be the pick of the crop so they last longer.
 
Emulex:

Thanks for the info -- that helps one understand what's going on. Still, if I followed the threads I was reading correctly, WD Cav Blacks come set up to keep trying, which apparently leads to the array dropping the drive.

So folks were using the WD utility to flip, say, Caviar Black .5 or 1 terr. drives and RAIDing them without issue -- they may not be as durable as the binned server models, but they are basically the same drives it seems, and not dropping the drives constantly made RAID setups work in desktop PCs for the right price.

Now, WD has apparently affirmatively made that impossible with a firmware update. This sort of thing is generally pretty annoying, unless it's really necessary to prevent problems.

Larry:

You are correct about WHS. I run one, and like it, and MS deliberately set it up to stay as far away from RAID as possible. Indeed, on the WHS forums, it's an almost pathological aversion sometimes. RAID isn't THAT hard (unless, of course, people start modding firmware in ways that make it so . . . )

I like the way MS set up the "pool" storage in WHS, but the irony is that you CAN also run a RAID array in the same system (just not as part of the pool, on grounds that such a thing would be silly). I do that with an old NetCell RAID 3 configuration, and it works quite well. And, as may become more relevant, it appears that non-pool storage may not be subject to the 2TB limit in the current gen WHS. Much of which doesn't matter if you're running a teeny box server. But mine is a home-built contraption with plenty of room for drives, and I had the RAID card and setup in the house. Space is good, so in they went.

I get fault tolerance with folder duplication on the "pool" side, and with mirroring on the RAID side. Still, I wouldn't generally think you'd go looking for RAID on the WHS side -- the basic setup is fine for what it's meant to do, and it's not like you'll have large number of users pounding the thing at home.

Actually, it's funny you should ask about WHS, and I was looking at a pair of the Caviar black 500s as additional pool drives. But these things get moved around from time to time, and knowing I couldn't use them in a RAID unless I hurry up and get the ones with the old firmware is what got me started on the question.
 
TLER is not a feature that has value for consumer RAID arrays. In the case of your RAID array, you want a drive that experiences an error to drop out of the array, because that drive has failed and needs to be replaced. TLER is designed for Enterprise RAID arrays where the performance loss from dropping an array into degraded mode while they wait for a replacement drive or rebuild to a hot-spare is unacceptable, outside of a maintenance window. For example, if you have a three-drive RAID5 array holding 2TB of porn and one of the drives dies, you're not going to care, just RMA the drive and rebuild the array when the replacement gets there. If you're running some 10-drive 18TB RAID50 array that's a file server for your company, you're going to hate your life if the degraded array can't keep up with demand during peak business hours, and you're going to want to be able to put the rebuild off until 10PM or something. (Examples chosen for clarity only)

I'd strongly recommend reading the Google Labs paper "Failure Trends in a Large Disk Drive Population", as it contains groundbreaking research about how and why hard disks fail that completely contradicted a lot of conventional wisdom. One of the major findings was that drives operate error-free until they begin to fail. It was previously assumed that a good drive could accumulate errors at some slow rate over its life, but it turns out that when a drive gets its first error, it will rapidly accumulate more until it fails completely. The upshot is that if a drive logs a single SMART error (even if it doesn't trigger the warning threshhold), that drive should be considered to have failed and be replaced immediately. (The one exception are link CRC errors, which can be caused by electromagnetic interference or power anomalies.)

Anyway, the reason I made this post was to reassure you that you don't need to worry that much about TLER. It will only come into play in the event one of your drives is failing, and the redundancy of your RAID array exists to protect you from drive failures. Of course RAID0 arrays would be one place where TLER make sense, given that it may potentially limit data loss to a single RAID stripe, but the better solution is to strongly reconsider whether you want to use RAID0, or just accept the risk and make frequent backups.
 
LtGR: Thanks. The example (colorful to be sure) illustrates the idea clearly, and I see your point. I'll take a look at the paper sometime, but that basic additional wisdom must have crept out, because I recall reading that somewhere else -- can't recall where.

The only remaining question would be whether there's something up with the WDs, or whether folks are making a change that's likely to hurt them (the problem with "consumer reviews" level wisdom being that it tends to be either really clever, dramatically dangerous, or both). Either way, this suggests that the company means well, so I get that point. But why then are people finding that the drives get dropped from the arrays, and that you can flip the TLER state and they'd then work consistently in the same arrays? If Google is right, the drive in question should fail rather promptly, should it not?

When I get time, I'll run through some posts again over at Newegg, and I might give WD a call. Your analysis suggests that folks should be able to run the drives in RAID setups without fiddling with TLER -- which would be fine. In which case, I might grab a couple. Indeed, this all has me pondering whether I might do a poor man's SSD by RAIDing the system for the next Win 7 build (my desktop). Even one of these new drives would be faster than my older (pre Veloci-) Raptor. Not sure whether the RAID would be worth the trouble, but WHS is there with the safety net, so it's worth pondering. And that seems a tad LESS esoteric than short stroking the drives.

Which leaves me back in the search for something on safe performance tweaks and their relative effectiveness on the desktop. Aligning drives with Diskpar or Diskpart is new to me, for example, though apparently mundane if you run an exchange server. I managed it, and the desktop PC where I did it prior to a Win 7 build does seem snappy. But there are enough other variables (new OS, +1 gig ram) that I don't know whether I've found a worthwhile tweak or whether I'm just in the stupid pet tricks category.

Thanks again for the post.
 
Last edited:
*Ahem* Um, well, I uh, take back any comment I made about short stroking being "esoteric." Maybe I should do a bit of homework first. If these figures are correct, this will get you a poor man's SSD for a price well within the beer budget.

The throughput looks good in general. Short stroked throughput is eye watering.

http://www.overclock.net/hard-drives-storage/618485-raid0-spinpoint-f3-hdtune-results.html

All that and space too for $100 (2x 500gb) or $180 2x 1TB). Pretty hard to resist. So I wonder if I can possibly hope to align and short stroke them. After that, I think I'd just give up and buy the SSD.

Cheers,
 
you have to consider fragmentation and latency - lets say you let your pc fragment for a week - you have typical # of files on o/s drive - run a full disk antivirus scan on short stroke versus an SSD drive. FF/IE cache/internet history/tons of little files here can multiply the fragmentation and latency problems with a cheap raid controller.
 
Yep, and further study suggests that RAID is generating most of the throughput, and that short stroking just keeps it consistent by not allowing the heads to travel inward. Worse yet, still further quick study in a review of the subject here suggests that all that eye watering throughput RAID generates doesn't help at all on the desktop.

Since I've seen the inconvenience of the array dropping a drive in the netcell array (bad usb cables as it turned out), which would not compare to the inconvenience if the RAID 0 dropped a drive with the OS running (even with WHS backing it all up), I'll just chalk this little thought experiment up to SSD envy and stick with a fast single drive for the next OS (forthcoming Win 7 install makes this the opportune time to buy one). Likely a WD Black or Spinpoint F3 1TB. Might nab a spinpoint or two for the server too -- it's surprising how fast those drives fill up, so adding storage there up to the 2TB limit may be worth a few bucks.

Ah well, the poor man's SSD idea was a nice fantasy whilst it lasted . . . thanks to all for the thoughtful comments . . .
 
Back
Top