Rules of thumb
Sep chaining is idiot-proof but wastes space…
Linear probing uses space better, is fast when tables are sparse, interacts well with paging
Double hashing is very space efficient, quite fast (get initial hash and increment at the same time), needs careful implementation, …
For average cost t
- Max load for LP (1-1/sqrt(t))
- Max load for DH (1-1/t)