> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Simon George
>
> I saw that plot in the CERN Tier0 talk at GridPP27, but the end of life
> failures were purely hypothetical as it only showed 3 years of data.
>
Ah yes, that was it. Or rather this was it:
http://www.gridpp.ac.uk/gridpp27/Tier0GridPP27.pptx
apparently the very neat black and white bath tub chart
comes from a paper published by Quantum[1] in 1999, at which
time the 'wear out period' as failure rates started heading
back up was at about five to seven years, whereas the CERN
stats fairly clearly show the beginnings of that at three.
I think Martin's presented similar figures from RAL, but
there were a lot of confounding factors from the well known
batch specific issues.
Ewan
[1] They used to make hard disks, and just about still were
at the time.
|