Comment by monocasa
4 hours ago
I mean, if you're a storage business, hopefully you've designed your architecture such that you assume drives will go bad, so you characterize the models of drive to make sure that not all the copies are on one manufacturer, and then you can take liberties finding the cheapest storage on the market. This only comes back to bite you when you didn't account for (because you didn't know) that there was decreased longevity, so your TCO calculation was off and you might not make as much money.
I agree with this.
However, an even more fundamental philosophy behind any work that we do is "defense in depth" which means that even after building the fault-tolerant, anti-fragile system, we also spend time and resources qualifying the inputs ...
... and then spend time and resources monitoring the outputs (error rates, failures, correlation, etc.).
Any one of those pieces is, theoretically, sufficient. Layering the pieces in a defense in depth strategy is what gives us the highly confident posture we enjoy.