I thought, oh, you only need two bits for the multiplier, that gives you four multiples of each tier size, that should be enough. But that means the largest bu

I thought, oh, you only need two bits for the multiplier, that gives you four multiples of each tier size, that should be enough. But that means the largest buffer (with a slightly less sophisticated formula than yours) is ((2 << 2) * 2 << (2 <<(8 - 2) - 1)) = 32 exabytes. Meanwhile, yours has a maximum of 60 GB. So i don't think you have too much granularity!