Slashdot Mirror
A Look Under Western Digital's Hood
Tom's Hardware got a rare opportunity to explore the Western Digital campus and show us what goes on under the hood of one of the favorites in storage tech. "When you buy a car, you look under the hood. Given the critical importance of hard disk storage in all of our lives, we thought you might want a peek under that hood, too. Now that Western Digital is in the business of breaking new capacity records (the latest Caviar Green was the first drive to hit 2TB, for example), we jumped at the chance to take a first-ever, unrestricted tour of its California R&D facilities. This is the place where magnetic technology of the 1950s meets the nano- and quantum-level technologies of the current decade."
and (my personal favorite) the warning about how the strong magnet inside the system could fritz your pacemaker.
So you have a large number of workers exposed to this machine that (I presume) creates massive electromagnetic fields? And they are exposed to it for lengthy amounts of time in proximity to it? And you have other workers in the same area/facility that are not exposed to it?
...
I tire of the ongoing debate that electromagnetic fields are hazardous to your health. Since you provide these people ongoing health care, perhaps you could release anonymized data so we could either confirm or deny this? If anything it would help clear things up in -- not only the power lines debate -- but also maybe cellphones if the EMFs are in anyway similar.
Just a thought
My work here is dung.
> Given the critical importance of hard disk storage in all of our lives, we thought you might want a peek under that hood, too.
Careful... what if you open the hard disk and there's a mouse inside who's been remembering it all for you? Then he sees you and realizes "Oh, crap, we need to do the experiment all over. Vogons! Oh, Vogons! Where are you...."
And then. Will come. The poetry.
-- IANAL, this isn't legal advice, and definitely isn't legal advice for you. Also, Squee!
I'm not sure about comparative reliability, but most WD drives come with a 5 year warranty nowadays, and their RMA process is the easiest i've seen. (I work at a repair shop, so we see a LOT of bad drives)
Well, all hard drives can fail sooner or later, and there's a reason for the M in MTBF. The problem with IBM Deathstars wasn't just that they failed (all do), but that their failure rate was disproportionately higher than any other brand at the time. And yeah, I had one of those fail on me too.
That said, I don't seem to have much of a problem with failing WD drives. I have a Raptor of each of the 75 GB, 150 GB and 300 GB varieties, all of them since that particular series was launched and all three still seem to chug along just fine. But that's a non-representative sample too, so don't take it as more than a personal anecdote.
A polar bear is a cartesian bear after a coordinate transform.
This article [>]
Ubuntu is an African word meaning 'I can't configure Debian'
Unfortunately, it is spanned [>]
Ubuntu is an African word meaning 'I can't configure Debian'
Thus, I won't read it unless someone provides a print-link (please?).
p.s. Sorry... Tom's hardware really annoys me, I just felt I had to do it.
Ubuntu is an African word meaning 'I can't configure Debian'
I worked in the Texas Instruments semiconductor fab shop in Sherman, TX for several years. Same sort of setup, different substrate (plus they don't have any etching processes). The bunny suits can get hot, but the sweat under the gloves make some work almost impossible. Try changing the battery in your watch wearing those plastic gloves and you'll see what I mean. Sometimes the gloves just have to come off; then you have to clean the work area thoroughly to decontaminate it (sodium in sweat was the biggest worry). One thing I'm curious about: vibration. We were in north Texas, and needed quite a bit of vibration control, mostly isolation pads. The article doesn't say where the WD facility is, I assume California. I see some isolation pads under equipment, but how do you handle vibration in a seismically active area?
Chaos maximizes locally around me.
The rash of 40GB HD failures (and it wasn't just W.D.) wasn't the hardware. In most cases it was because the entire 40GB was partitioned as One Big Drive... in FAT32, which was still the dominant filesystem for Windows.
The problem is that FAT32 has a bug that can cause data-wrapping if the partition is larger than 32GB. And the bug exactly mimics a failing HD -- random data loss, corrupted files.
The explanation used to be on Microsoft's tech site, but it vanished last time they nuked a bunch of older material (which they do periodically).
At any rate, you can see why there was a rash of "HD failures" when HDs exceeded 32GB. And W.D. took the brunt of this, since at the time they were the first (and for some time, the only) manufacturer offering a consumer HD larger than 32GB. By the time everyone else caught up, most of the Windows world had moved to NTFS, which does not have the bug, and the problem went away.
BTW not long after that, Seagate did a study on RMA'd drives, and found that about 60% of the time the hardware was fine, and the "failure" was in fact caused by a filesystem or software error. This is pretty much in line with my own experience.
~REZ~ #43301. Who'd fake being me anyway?
If I were you I'd read the article to the end before writing. The last few photos show their SSD labs. Apparently they are well aware of the situation.
I'm sure this is purely anecdotal, but it's been my experience with WD drives that they either fail spectacularly within the first few months of operation or I can run them into the dirt over the course of 8+ years without a hiccup. Also, I've had nothing but fantastic experiences with their warranty processing department - their RMA program is quick and painless should I find myself with a dead drive still under warranty.
As an aside, as a SAN administrator I feel it important to point out that regardless who manufactures the drive: It's a matter of when, not if, the drive will fail. If the data retained on the drive is critical (business or personal), one must always be prepared for the eventual death of the drive and plan accordingly.
I see your Seagate anecdote with one of my own.
I had a Seagate 500GB 7200.10 fail in September 2008 (crappy firmware edition), not long after installation after having sat on a shelf for a few months. When I approached Seagate to RMA the drive, they barely bothered to ask me what was wrong with it. Filled out a form, slapped it into a box, and back came the replacement, though a little less promptly than 3-7 days on the RMA form.
I'm at the point in life where I generally install a new OS onto a new disk drive, re-using older drives after *months* of successful operation on the new system / configuration. Spindles are cheap insurance at modern prices.
No vendor is immune from production glitches. I've been searching for the fountain of electronic youth for twenty years. No company, however great, is immune from a Toyota moment.
It amuses me how talismanic we tend to become on low sample sizes. Typical example: "I had a Brand X drive fail on me back in 2000, and haven't purchased another one since." Every vendor I've ever tried has fallen on its platter at least once, so I'm now back to pencil and paper.
What would make me happy is more binning from the drive manufacturer's. I like the middle bin between Joe consumer and Enterprise exabucks. It can't be that hard to look at production data and say "this batch is better than that batch" and bin accordingly.
I've heard that the external backup drives sold at Costco and places like that *are* sourced from batches not up to full warranty treatment. You'll notice these appliances have a shorter drive-life warranty than the same drive sold naked.
OTOH, it's hard for the average consumer to know for certain if you pay a $50 premium for the extra quality bin whether you're getting more quality, or just a different sticker. A web hosting facility is going to have the failure data to back up any decision making on paying a premium price.
Also, it's pretty easy for a careless consumer to compromise drive life by poor handling, installation, or faulty cooling. I'd guess about half of all failed drives (excepting DeathStar production sagas) suffered abuse at the hands of the retail chain or end user, which sets the limit on how much quality it makes sense for the vendors to promise.
However, if the consumer is playing a $50 sticker premium for a "black edition" disk drive, it's also likely the abuse level and cooling components are more carefully considers.
That would be a funny business model. The drive vendor sells exactly the same drives for $50 more, but the buyers who spring for the premium take so much better care of the drives, the drives gain a reputation for delivering higher reliability justifying the price.
To make this work, the vendor has to keep the supply of "black edition" drives to a relatively small trickle. Once the masses get their hands on them, the game is ruined.
One point the article doesn't mention is analyzing the platters under static load instead of dynamic load (including strain from spinning) and spindle vibration. I wonder how much that complicates quality control.
Western Digital does one better - they will cross-ship a new drive if you need for only the cost of your return shipping to them. It's really handy when you have a drive kick up a pre-fail SMART error. You can get a new drive on the way before the old one fails and just do the swap out in the array. I've had to do it twice (out of 16 drives) for my home storage array over the years.
The thing that cheesed me off most about the whole ordeal was that they issued a firmware "fix" that bricked the drives outright.
At that point I started to suspect their in-house testing checklist looked something like this:
1. Compile it.
2. Release to the customer.
3. ???
4. Profit!
I mean, seriously, it's a stinker of a bug, but there's a step missing between 1. and 2.: "Get a few hundred drives from the warehouse, do random number of R/Ws, image, set up for failure. Test to see if bug is fixed, also test for bricking / regressions / other issues". Screw the cost, get the engineers some drives from the warehouse, get a few from RMA that have failed, and let them do some testing.
The clincher was that the first firmware update didn't fix the whole issue: while the bricking could be considered a bigger problem, the update still didn't fix the bug -- you could get past the "bricking" with the serial console, but the drive would still crap itself when it saw the trashed SMART log record.
As for the whole RMA procedure, they made a colossal clusterfsck of it. The front-line staff didn't know a thing about the bug (even though it was on the knowledge base), and just played the "stonewalling game". As in, "it's a problem with your hardware, the drive is spinning so it's fine."
As was, the CSRs didn't know anything about the "firmware issue" (Seagate refused to call it a recall) until near the end, and SG themselves just kept making fuckup after fuckup until it all ballooned into one giant clusterfuck.
Given that they had their own in-house data recovery service, and that they knew how these drives were failing, they should have (at the VERY LEAST) offered to repair them free-of-charge regardless of warranty status. It's a firmware bug, thus it's Seagate's fault.
It seems a lot of "customer first" type policies have fallen by the wayside recently... Now it's pretty much "take the customer for all they're worth, and hope they don't tell their friends/the cops that we were naughty."
My opinion of Seagate was soured before the 7200.11 issues though -- I bought a 500GB 7200.10, which died within about 8 months. Basically, the motor locked up mid-spin, and (AFAICT) the motor control chip decided to slam on the brakes (short all 3 motor coils to ground -- aka dynamic braking). Big mistake. The drive launched itself across the floor (the cables were pretty loose) and nailed the side of my leg. It wasn't especially painful, but certainly brought me back into the "real world" (I was in the middle of a huge mess of coding).
The next morning I called Seagate, spoke to a really apathetic CSR who spoke to me like I was interrupting something far more important, and who couldn't give a flying crap in a storm about issuing an RMA number. During the 15-minute call, the CSR outright refused to escalate the call to a supervisor ("we have no supervisors here"), and just kept giving me the same answer time-and-again.
I gave up and called the company that sold me the drive (CCL Computers in Bradford). In 5 minutes I had an RMA number, and instructions for returning it. "It might take a week to get it tested, but we'll replace it if it's faulty."
Two days later I had a new drive sitting on my desk at work. Now *that* is customer service.