[Beowulf] Are disk MTBF ratings at all useful?

Ellis H. Wilson III ellis at cse.psu.edu
Thu Apr 18 18:53:46 PDT 2013


On 04/18/2013 07:01 PM, mathog wrote:
> How do they come up with the MTBF values for disks anyway?  Clearly it
> is not based on watching a large
> sample of disks for countless years!

I am not intimately familiar with how they come up with the values (else 
I probably would be at liberty to comment), but I do know there is an 
increasing diversity even within the same manufacturer and even the same 
batch, so my short answer to the subject is "not really."

A really great paper that examines the performance side of this, by a 
guy I had the privilege of sharing a room with two CMU Parallel Data Lab 
retreats ago:

www.usenix.org/event/hotos/tech/final_files/Krevat.pdf

Again, this paper is only regarding the performance side, not MTBF, but 
I would expect (from only a superficial, perhaps wholly incorrect 
standpoint) that this would correlate to the failures aspect as well. 
Just me speculating here.

Best,

ellis


More information about the Beowulf mailing list