LWN.net Logo

KHB: Real-world disk failure rates: surprises, surprises, and more surprises

KHB: Real-world disk failure rates: surprises, surprises, and more surprises

Posted Jun 14, 2007 4:16 UTC (Thu) by pr1268 (subscriber, #24648)
Parent article: KHB: Real-world disk failure rates: surprises, surprises, and more surprises

Thank you for the article - this gives some insight into how hard disk manufacturers can make claims for X hours MTTF and similar. And to make general estimates as to the life span of disk drives in general.

Veering slightly off-topic, I have become an exponent for Seagate hard drives - any company who proudly advertises "5 year warranty" on the packaging of their mass-market consumer IDE and SATA drives (whilst most other major manufacturers only give 1-3 years) gets my business. It's not necessarily about whether I'd really have to make a warranty claim (WDC honored such a claim for me back in 2002 with minimal fuss), but rather that they have such a high level of confidence in their craftsmanship as to even advertise such a warranty.

Sorry if this violates any rules for "plugging" a particular brand. But, I'm reminded of what the salesperson at Fry's told me about why Seagate warranties their drives so well: They re-engineered the spindle bearings and motor assembly - the critical points of the disk drive which the salesperson said were most often the cause of total drive failure. Any other ideas/comments?


(Log in to post comments)

KHB: Real-world disk failure rates: surprises, surprises, and more surprises

Posted Jun 14, 2007 7:03 UTC (Thu) by bronson (subscriber, #4806) [Link]

Er, you listen to salespeople at Fry's??

I've only ever managed clusters of disks numbered in the few tens. Overall I find that hard disks are surprisingly reliable. I think most disks fall out of rotation because of a lack of capacity, not because they break. (I still remember how thrilled I was when they fit 20 GB of data in a 3.5" package... today I can fit 50 of those drives into a single 3.5" package!)

One lesson that I've learned is that manufacturer loyalty is pretty much meaningless. I had an early batch of Quantum 15G drives that were so stone-cold reliable I'm sure they would still be working today. However, Quantum Fireballs would reliably die after two years. I remember Maxtor producing utter crap in the past but I have a set of their 60GB drives still spinning. Seagate used to be fairly mediocre and now they're top notch across the board. When deciding what drive to buy, go to Storage Review and read about the individual models; brand is meaningless.

I also find that motherboard failures (i.e. CPU or memory socket corrosion, weak power supply, etc) tends to destroy the drive as well. I read the Google paper back in Feb and I don't remember them taking this into account... It would be nice to know what portion of failures were drive-only, and what portion was chaotic damage that just happened to include the drive as well.

KHB: Real-world disk failure rates: surprises, surprises, and more surprises

Posted Jun 15, 2007 0:51 UTC (Fri) by pr1268 (subscriber, #24648) [Link]

> Er, you listen to salespeople at Fry's??

I listen to them, but I don't always believe everything they say. But, again, my Seagate discussion was more tuned into the idea that their warranty (which they proudly advertise in big print on the sides of the box) is much longer than those of most other name-brand hard disk manufacturers (which is often hidden in the fine print). I agree, brand loyalty doesn't mean much these days, and besides, I used to think WDC (Western Digital) drives were the finest-quality consumer hard disks. Nowadays I don't think WDC drives are nearly as good as they were 5-10 years ago, and I've never had any good experiences with Maxtor. Times change, and so do hard drive manufacturers' quality control (and standards).

Maybe in 5-10 years I'll think the same about Maxtor, WDC, or Fujitsu as I do now about Seagate...

KHB: Real-world disk failure rates: surprises, surprises, and more surprises

Posted Jun 15, 2007 4:13 UTC (Fri) by njs (guest, #40338) [Link]

It's a funny thing about us humans, if you put us in a data vacuum, we find some way to fill it, whether it makes sense or not. Most of us have no useful statistics at all on what hard drive brands or models will be reliable; heck, as the article points out, the people *making* the drives don't even know this.

But this ignorance makes us so *uncomfortable* that everyone finds some random fact to base their decisions on, like an anecdote about that time they bought a *** brand drive they and it died after a week and so they never buy *** anymore, or how they heard the new *** brand drives use a fancier production process, or the warranty labels on the side of the box.

Warranties aren't a measure of how proud some engineers somewhere are, they're a measure of some sales/accounting decision about how the cost of providing that warranty will compare to the extra sales they get by putting it on the box. (5 years ago we were using what, 40 GB drives? If that died today, assuming you even still had it online, would you figure out how to ship it back for a new 40 GB under the warranty, or just pick up 400 GB at the local mall? Whenever it's the latter, the warranty costs Seagate nothing, and Seagate knows how many people fall into each camp.)

Hard drives are a commodity. Any given model has some greater-than-zero failure rate; people who care about their data make backups and the failure rate doesn't matter, people who don't care about their data worry and fret over exactly what the best lottery ticket to buy is. Me, I figure hard drives are all close enough in speed I'm never going to notice, but I have the thing sitting right next to me all day long, so I buy drives by checking Silent PC Review's recommended list, and picking the top-rated drive I can find on sale.

Hard Disk Drive Warranties

Posted Jun 15, 2007 5:06 UTC (Fri) by pr1268 (subscriber, #24648) [Link]

Even assuming that hard drive warranties are written by the sales/accounting department, don't you suppose that they looked at return rates of their products in order to make that warranty period?

WDC used to pledge a 3-year warranty. Now it's 1-year (again, assuming their consumer drives--IIRC their "Raptor" series of true-SCSI drives gets a longer warranty). Whether it was the sales/marketing folks at WDC, or it was the engineers, either way, around 3-4 years ago they decided that the warranty claim rate wasn't good enough to justify maintaining the 3-year warranty, so they reduced it to 1-year.

Certainly the folks over at Seagate were wise enough to perform the same cost vs. benefit analysis of pledging such a long warranty, regardless of whether it was the engineering team or the sales/marketing folks. But, with Seagate's substantially longer warranty, I can only assume that their cost vs. benefit analysis demonstrated either of two things: (1) their drives were high-enough quality such that the return rates were low and they could warranty their drives for 5 years whilst remaining profitable, or (2) They could absorb the cost of replacing defective drives under warranty at will for the indicated warranty period given a failure rate no better or worse than the commodity average.

I just don't see (2) above happening without Seagate making drives of such sorry quality and cheap manufacturing costs that they can justify the long warranty (analogy: I sell you a television for $150 which cost me $20 to build, and it has a 20% annual failure rate, so I can justify warrantying it for 5 years and still make a profit of $50), and I don't see them making drives of such unusually high quality that their manufacturing costs (and retail prices) spiral upwards. Their drives are competitively priced with WDC, Maxtor, and Fujitsu.

I don't mean to argue; but rather, I wanted to share my experiences and perhaps invoke a mildly-stimulating discussion. I totally agree that doing some basic consumer research on hard drive quality and features (you mentioned Silent PC) is a good idea for anyone wanting to invest in spinning platter data storage. :-)

Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds