Calling All Techies: Hard Drive Question

[Guest rubidium]

Advertisement (gone after registration)

As a scientist, it's hard for me to sit idle and read a lot of anecdotal commentary on the poor reliability of hard drives. The fact is that the "big four" - Western Digital, Hitachi, Maxtor, and Seagate - are all very competitive. The WD drives that I've been using for years have an MTBF of 1.2 million hours. That is, if operated continuously, the probability of failure is as follows:

 

Years Prob of Failure

0 0.0%

1 0.7%

2 1.4%

3 2.2%

4 2.9%

5 3.6%

6 4.3%

7 5.0%

8 5.7%

9 6.3%

10 7.0%

11 7.7%

12 8.4%

13 9.0%

14 9.7%

15 10.3%

 

If that's not good enough, then the reliability using 2 redundant drives - either as RAID1, or frequent backups of one to the other - is quite impressive:

 

Years Prob of Failure

0 0.00%

1 0.01%

2 0.02%

3 0.05%

4 0.08%

5 0.13%

6 0.18%

7 0.25%

8 0.32%

9 0.40%

10 0.49%

11 0.59%

12 0.70%

13 0.82%

14 0.94%

15 1.07%

 

The problem is that most hard drives are placed in operating environments that have limited air flow, and thus run at excessive temperatures, thereby lowering their reliability dramatically. Open up any mass-produced PC and you'll see that conditions are not conducive to supporting high reliability with any component. (Why should they be? Everything is done as cheaply as possible. Also, it's not good business to build a computer that lasts for 10-15 years.) The same goes for most external hard drives. The only company I know of that goes out of their way to offer external hard drives having enclosures built (using heat sinks and good ventillation) to realize the long design life of the drives within is G-Technology Inc. However, their prices reflect it.

 

Jim

[marknorton]

Even if you believe the design MTBF figures, arrived at by accelerated testing, that still represents a failure rate of 1 drive in 137 per year and my own experience is that they fail more frequently that this, especially laptop drives.

 

Whether it's down to temperature, disk activity, power on time, temperature cycles, load/unload cycles is open to debate. The Google article suggests that drive failure rates are not well correlated with temperature or actual disk usage.

 

Most drives will be replaced before they reach the end of their working lives - I have a box full of old hard drives including 3 WD drives from 1995 with a boggling capacity of 120Mb. As technology moves on, we never get to see what the actual lifetime is but the small proportion failing in the meantime cause problems. What price your data?

[sean_reid]

As a scientist, it's hard for me to sit idle and read a lot of anecdotal commentary on the poor reliability of hard drives.

Jim

 

I would love to see data beyond the anecdotal info. (which I find useful nonetheless). Do you have links?

 

I ordered the WDs but still find this to be an interesting topic.

 

Cheers,

 

Sean

[ericperlberg]

I would love to see data beyond the anecdotal info. (which I find useful nonetheless). Do you have links?

 

I ordered the WDs but still find this to be an interesting topic.

 

Cheers,

 

Sean

 

It's here.

 

Quoting here:

1) Failure rates are known to be highly correlated with drive models, manufacturers and vintages [18]. Our results do not contradict this fact.

2) We do not show a breakdown of drives per manufacturer, model, or vintage due to the proprietary nature of these data.

 

What they do is examine a wide range of assumptions about what drive failure is and what causes it.

 

A report is out today on Ars Digita about a company out of Israel which has received strong funding from some very big IT players for a 5 terabyte storage medium disk essentially the size and shape of a DVD. Hopefully it will also be archival.

 

Good luck with your WDs Sean.

[sean_reid]

Thanks Eric and thanks for the link.Too bad they won't talk but the largest hard drive study done yet (to my knowledge) did indeed find differences among "drive models, manufacturers and vintages". So much for the idea that brand and model doesn't matter. Thanks for the legwork.

 

Cheers,

 

Sean

[marknorton]

Every manufacturer seems to have suffered from sudden death syndrome. That said, I and several friends who are also pro photographers have had good luck with Maxtor. A Western Digital in my last Dell failed miserably. I'm about to call Dell about adding a second internal hard drive - and hoping they have more than one choice

 

John I think that's it. People make the best informed choice they can but the drive they buy is just one point on a statistical distribution and their drive - which is all that matters to them - may turn out to be significantly better or worse than the average. Those averages are tough to ascertain - the Google study points to the MTTF figures being over-stated and ignoring infant mortality.

 

The best an individual can do is to try to keep ahead of any disk failures by taking backups and keeping an eye on drive performance.

 

Let's hope the girls in (say, Singapore) churning out Sean's WD drives (for say, $40 ex factory) had their act together on the day.

[dist]

Advertisement (gone after registration)

I have a couple of Hitachi flatliners to report. We manufacture mobile geophysical equipment, and some data recorders are fitted with 2.5" laptop HDs (yes I know that these are different beasts but then again they are supposed to be used in a mobile scenario). Having had a couple dozens of those out in the field I guess I can, to some extent, tell the good ones from the lemons. Back when Hitachi was still IBM, mortality was close to non existent (that kind of track record prompted us to pick IBM in the first place). Ever since IBM sold their HD business, quality never was the same again. We keep getting dead disk drives back from irate customers. We have switched to Fujitsu at a later time, those seem to be much sturdier.

I'm staying clear off Hitachis.

Cheers,

[bernd_muc]

RAID is a great way to go for automating your backup

NO!!! No RAID is a substitute for a backup. If you erase a file on mirrored disks, it still will be gone. Any backup needs to be on a seperate media which you copy your data to. This way, you'll always have all your files, even if one gets lost (for whatever reason). Now, finding a backup media for 2TB of data might be tricky (and/or expensive). You might just go ahead and get two of those RAIDs and copy the data from one to the other periodically (don't mirror them).

 

If you use a RAID, be sure to frequently check the S.M.A.R.T. status of the disks and replace any that isn't "healthy" as soon as possible. Also, the mentioned Google study says, that if a disk survives the first few weeks, it is likely not to fail to quickly.

 

Bernd

[wparsonsgisnet]

Actually, I understand that some drives were more reliable than others but Google, for various reasons, is keeping mum on that topic.

 

Cheers,

 

Sean

 

Yes, their article did allude to one quality problem related to a single drive, and they didn't talk about it.

 

The items I found most interesting in their article were:

1. drives should be left on,

2. if drives lasted more than 1 year, they would last longer than average, and

3. mostly, the consumer quality drives are about the same. This should not surprise us -- info is much too available on the web. Next, people will be using the web to tell Leica what they think.

[wparsonsgisnet]

... Let's hope the girls in (say, Singapore) churning out Sean's WD drives (for say, $40 ex factory) had their act together on the day.

 

Isn't it more like $4?