3 ms·
Using a 32 bit ID is an interesting choice. It means you can only index 64-bits per bucket. I wonder if using a varint encoding would give you even more savings
by vlovich123 4y ago
Using a 32 bit ID is an interesting choice. It means you can only index 64-bits per bucket. I wonder if using a varint encoding would give you even more savings while handling > 4 billion documents at the cost of a bit more expensive serialization/deserialization cost (which should be negligible in the grand scheme of everything else being done).