It seems odd to look at failure rate in isolation, without considering cost and density; at scale, improved cost and density can be converted to lower failure r

It seems odd to look at failure rate in isolation, without considering cost and density; at scale, improved cost and density can be converted to lower failure rates via more aggressive RAID redundancy, no?