Question regarding actual memory usage savings about bloom2 HOT 1 CLOSED

Hey Jimmy,

It works out to be ~1.5% - so if a fully loaded bloom filter (every bit set) was 100%, a fully loaded bloom2 instance takes 101.5%. Obviously the use case where the tradeoff makes sense is for partially filled bloom filter, where the 1.5% overhead is worth spending to gain a large reduction in total size proportional to the load factor.

I don't have any actual numbers on hand for load factor -> savings, as it's dependent on your hash and input data, but unless your heavily loading the bloom filter (which seems unlikely given the false probably rates) I would expect it to be a net saving. I'd always recommend benchmarking this (or any other library) for your own use case!

Dom

from bloom2.