Comments on: Google’s Disk Failure Experience

By: OCTO talks ! » Les Patterns des Grands du Web – L’obsession de la mesure

Sun, 20 May 2012 13:26:14 +0000

[...] [i] http://storagemojo.com/2007/02/19/googles-disk-failure-experience/ [...]

By: Big Data: Are there any examples of firms using predictive analytics for planning preventative maintenance (e.g IT equipment failures)? - Quora

Fri, 04 May 2012 18:13:20 +0000

[...] own data centers, most notably the report they published on hard drive failures (referenced here): http://storagemojo.com/2007/02/1…Comment Loading… • Post • 11:13am Add [...]

By: Offline uncorrectable sectors

Offline uncorrectable sectors — Sun, 08 Jan 2012 08:56:11 +0000

[...] weird when the hard disk is brand new and it was installed by my colleague. Not really, look at Google’s Disk Failure Experience. With 1-year AFR(Annualized failure rates) dropping significantly from the AFR observed in the [...]

By: Damaged SSD's? - Tech Support Forums - TechIMO.com

Damaged SSD's? - Tech Support Forums - TechIMO.com — Sat, 30 Jul 2011 01:26:16 +0000

[...] rate, in the first, few weeks, than than do for the next 2 years. Then it starts, to go up again. Google’s Disk Failure Experience I just don't see how the power supply, can run a high performance video card, and be bad. Nor can [...]

By: OCTO talks ! » Datacenter as a Computer : une plongée dans les datacenters des acteurs du cloud

Mon, 06 Jun 2011 05:40:45 +0000

[...] Storage. Stocker les données Plusieurs systèmes possibles : – du NAS. De mémoire, le papier ne s’attarde pas trop sur l’utilisation ou pas de SAN (Difficile donc de savoir s’ils en utilisent ou pas et si non, si cela est du à des coûts prohibitifs ou des fonctionnalités inadaptées à ces architectures distribuées) – du Distributed File System s’appuyant sur des disques locaux. Ainsi la résilience est gérée par le Distributed File System (et la réplication au niveau applicatif des fichiers) et non par l’infrastructure “hardware” sous-jacente. A titre d’exemple, c’est un axe trés fort des solutions “NoSQL” ou même “in-memory datagrid”. En termes de disques, ce sont plutôt des “desktop-class” que des “enterprise-grade” principalement à cause de la différence de prix. Ce n’est pas sans rappeler l’étude de Google sur les taux de pannes des disques durs. [...]

By: Ilari Halminen

Ilari Halminen — Wed, 27 Apr 2011 15:24:59 +0000

Well, I cannot figure how the temperature / error statics are made. It seems that drivers in cold temperature will fail easially.

I get an idea, that hard thing would be changing temperature, like that drivers are in cold and then get in use sometime warming up. And after that cooling down again for next loop.

Well what you do think about that?

By: AJC

AJC — Thu, 07 Apr 2011 13:06:36 +0000

MTBF is Mean Time Between Failures
The defintion described here is MTTF, Mean Time To Fail
MTBF describes the time between two failures, hence the word BETWEEN. This value is only a valid entity during the normal life span of a component, and not during Burn-In or Ware-Out, a so called Steady State variable, derived from the statistics
And the math is easy to simplify. If you n drives with MTBF of 300000 hours, then the MTBF of you complex would be 300000/n hrs.

By: KD Mann

KD Mann — Thu, 24 Feb 2011 18:08:13 +0000

When this paper came out, I (and a lot of others) gave it a lot of credence.

After having a look at how Google deploys disk drives, I reach the conclusion that this study is nearly meaningless. One cannot draw any conclusions whatsoever from reliability data taken from a pool of disk drives that are (literally) velcro taped to a rattling sheet of tin. Actually, it’s worse even than that…see the photos:

http://news.cnet.com/8301-1001_3-10209580-92.html

That google’s disk drive population performed as well as it did, under these ridiculous environmental conditions, seems a tribute to the robustness of the technology.

KD Mann

By: Backing up is critical: Part 1 « Learning Across Generations

Backing up is critical: Part 1 « Learning Across Generations — Fri, 11 Feb 2011 18:47:39 +0000

[...] rate of hard drives is probably higher than you think? Check out here for more information on the failure rate from Google who is the leader in hard drive [...]

By: Quora

Quora — Wed, 09 Feb 2011 04:12:11 +0000

Can a hard disk degrade if I format it a lot by installing and reinstalling Operating Systems?…

The more you use a hard drive, the sooner it will fail. However, I seriously doubt that you’re adding enough extra use to the drive to cause it to fail materially sooner. The definitive study on consumer-grade hard drives was actually done by Google a…

By: Robin Harris

Robin Harris — Mon, 31 Jan 2011 23:41:20 +0000

Rockmelon,

Boy, is my face red! You are absolutely correct. I’m surprised since the post has been up for years.

Anyway, I’ll update it. Thanks for the catch! And more later.

Robin

By: rockmelon

rockmelon — Mon, 31 Jan 2011 02:00:43 +0000

Hi Robin. I think you are out by a factor of 2 (roughly) and this misinformation is now being spread by the Wikipedia page for
Annualized Failure Rate, which links to and quotes from this
page.

Look how Seagate calculate AFR from MTBF at
http://enterprise.media.seagate.com/2010/04/inside-it-storage/diving-into-mtbf-and-afr-storage-reliability-specs-explained/

1 – e(-8760/300000) is an AFR of approx 2.88% so you should
expect 17267 drive failures annually from a population of 600000,
not 8760.

By: Fascist Nation

Fascist Nation — Sun, 16 Jan 2011 14:00:12 +0000

I was aware of Google’s paper since it first became available and have re-read it several times when I had a point to make to someone about drive reliability. But having just run across your site from a Google search for some related matter, I must say I had some serious belly laughs over your synopsis of this paper. Thank you for the insight and keep up the humor.

“Vendors define failure differently than you and I do. And, oddly enough, their definition makes drives look more reliable than what you and I see.” Classic!

By: MTBF - bill’s blog

MTBF - bill’s blog — Fri, 04 Jun 2010 03:41:42 +0000

[...] R., (2007, February, 19th), Google’s Disk Failure Experience, retrieved on June 3rd 2010 from http://storagemojo.com/2007/02/19/googles-disk-failure-experience/ My Life, School, System Administration none Edit in [...]