Hacker News Reader

Comment by SixSigma | Hacker News Reader

A useful additional metric is the age of the drive at failure.

This would determine if the failure rate was constant for the life of the drive (meaning random failure) or is it age related (infant mortality or old age).

25 drives that fail after 1 week plus 25 that fail after 50 weeks is different to 50 drives that fail one per week.

View on HN

Luckily they open source their operational and smart status for all of their drives[0]. This means that you can do this additional analysis (and more). Which is awesome.

[0] - https://www.backblaze.com/hard-drive-test-data.html

Brilliant, thanks for letting me know. I'm studying Logistics Engineering and reliability analysis is part of my degree. This will make a great case study :)

To save you some time, they did this analysis previously. I can't find it, but the summary was that the failures happen at the two ends and not a lot in the middle. ie. a bunch die early (infant mortality) and the rest die pretty late (old age).

This was the original blog post where we analyzed hard drive reliability and saw that it followed a bathtub curve: https://www.backblaze.com/blog/how-long-do-disk-drives-last/

Yep, that's the one I was thinking of. Curious, did you ever do a follow up? Have 50% of the original drives died yet?

Well, that's just the analysis I was thinking about. Good looking charts there, nice work.

Reply on news.ycombinator.com