linuxhw / smart Goto Github PK

The current calculation of MTBF seems to unfairly place an emphasis on drives that only have a few golden samples, with the top drive of each drive category each only having 2 samples each - with similar results when it is sorted by vendor.

I believe that using cumulative drive days to calculate an Annualised Failure Rate, like Backblaze, would deliver results that are more representative of their real world reliability, as it would place a larger emphasis on larger sample sets with fewer failures.

One issue however is drive duplication, as I don't believe that the serial numbers of drive probes are stored? I guess it could be deduplicated by only considering the last probe of a drive for a specific HWID or Inventory ID but I'm not sure how the HWIDs are generated so I don't know how persistent they are with hardware changes or if individual devices have their own calculated HWID.

Backblaze appears to use the following formula to calculate AFR[1]:
AFR = ( Drive Failures / ( Drive Days / 365 ) )

Hi, I just attempted to sudo -E hw-probe -all -upload on openSUSE Tumbleweed. Do you know how to correct this issue?

-Thanks