[TriLUG] Compliments to David Zeuthen for Disk Utility

[John Brier]

On Thu, Sep 16, 2010 at 2:28 AM, Aaron Joyner <aaron at joyner.ws> wrote:
> Google has done a lot of work on attempting to predict hard drive failure
> rates.  Based on the published research, it's true that reallocation counts
> do correlate with an increase in average failure rate of drives.  A quote
> from the paper, "After their first reallocation, drives are over 14 times
> more likely to fail within 60 days than drives without reallocation
> counts...".  There are some other SMART signals that are interesting as
> well, such as scan errors, offline reallocation counts, and probational
> counts.  Unfortunately, part of the core conclusion is that "Despite this
> high correlation, we conclude that models based on SMART parameters alone
> are unlikely to be useful for predicting individual drive failures."  That
> said, it's one component that should probably warrant closer scrutiny of
> that disk and it's performance and behavior in the future.  For more details
> and a really interesting read:
>
> http://labs.google.com/papers/disk_failures.html

I remember someone linking that paper to me a while back. I didn't
remember those quotes, but that adds to my understanding of the
significance (or lack of significance) of reallocated sectors.

Thanks, it seems to be true in my case too, because I've had a single
reallocation on that drive for at least a year, and no problems since
then.

JB