I don't understand the hash index part. I guess that for every segment on disk, you also have a hash table for it, correct? Also, this part doesn't seem to be very memory-efficient