... which is 11x larger than bs'. Nevertheless, we are much faster cold for 0-10 entries, very respectable for middle tier, and significantly faster for very large sets. So I figured it isn't worth the extra 200GB or so. Fulcrum can also win on the largest sets with an unbounded cache setting, this shows its installed default (as it shows our default hashmap load factors, which target a 32GB machine).
(edited)