3

When looking for large new drives (>= 1 TB) on newegg and the like, I note a number of reviews talking about drives being either D.O.A. or hitting the Click of Death (or even releasing the Magic Smoke) within a week or so of use. A portion of the reviews mention this phenomenon whether the drive in question is Western Digital, Hitachi or whatever.

For those of you using Windows, what do you to:

1) Place a large initial stress on the drive to see if it can take it? For how long?

2) Test the drive afterwards (presumably with some sort of S.M.A.R.T. tool or others) to see if any negative changes have been noted?

Note: This is one component of a larger plan for both high-availability and backups for my home data.

Improve this question

3 Answers 3

Reset to default
5

Current Mean Time Between Failures(MTBF) for consumer SATA drives is 10^14 write bits. This means that if you have a 2TB drive and write every bit 50 times on the drive, and then try to read every bit, you will have a disk failure in that time.
Google has done a lot of statistical analysis of consumer disks and have noted several trends, the most important to your question is that you will get about a 10% chance of failure in the first 3 months if you put your disk in high-utilization (stress test). If it makes it past the first 3-6 months, then it will last 3-4 years before the failure rates starts going back up again.
It's interesting reading if you like statistics, and even those of us who don't, still get the idea from the graphs...
Google link
Another source is Carnegie Mellon

EDIT: One other thing that relates to your question from the Google paper, is that these rates apply across all drive manufacturers. Google buys whatever gives them the most MB for the buck.

Improve this answer
5
  • I've read the paper. The paper is what prompted my question about stress testing, and the conclusion that it is down to batches within particular models, rather than drive manufacturers, is why I was keeping the question manufacturer agnostic by noting that the phenomenon appeared over all drives.
    – MetaHyperBolic
    Dec 15, 2009 at 20:45
  • Um..that 10^14 number is used a bit out of context here. The 10^14 bits is the non recoverable bit error rate...which is where the bit flips undetected on the disk resulting in an unreadable sector. It's not a disk failure. But would corrupt the file in question assuming you aren't running any sort of RAID to be able to deal with the error.
    – 3dinfluence
    Dec 15, 2009 at 21:00
  • @3dinfluence: Depends on your definition of a drive failure. The Carnegie Mellon paper brings this question up: What is the definition of a drive failure? To a manufacturer, the silent write issue (bit-flip) may not be, to the consumer, it is 100% of the time. I have never run into anyone who would say that the drive lost my data, but that it isn't failed. Long and short, I am not convinced that 10^14 is out of context...
    – Scott Lundberg
    Dec 15, 2009 at 21:37
  • 1
    @Scott Lundberg: Well the question is about drive failure not drive errors. This is an error, no doubt about that. But there's no way to avoid this error you can only try to minimize your risk exposure due to it. Statistically you're just as likely to run into this error on any read from the drive. It has nothing to do with if the drive is healthy or not. You would even be able to write data back to the same spot on the drive and read it back successfully. The real danger here is running into this error during a rebuild of a degraded RAID5. blogs.zdnet.com/storage/?p=162
    – 3dinfluence
    Dec 15, 2009 at 22:57
  • Fair enough. Guess we will have to leave it to the OP to decide what he feels is a failure. :-)
    – Scott Lundberg
    Dec 16, 2009 at 0:28
4

I don't stress test them at all but I do keep an eye on SMART values. I use Speedfan or HDtune to view the smart data.

Improve this answer
1
1

Spinrite does this, both to recover lost data and also as preventative. Using mode 4, it flips each bit 4 times (on/off/on/off or vice versa) looking for bad sectors. If it finds them, it attempts to correct them and mark that sector as bad, or it moves on. Great maintenance for drives and forcing SMART data

Improve this answer

You must log in to answer this question.