Functional bloom filter, better than hash tables

Abstract

Hash tables have been widely used in many applications, which need to return values corresponding to each input key. However, hash-based data structures have an intrinsic problem of collision, where different keys have the same index of a hash table. As the load factor of the hash table increases, the number of collisions increases. Elements that could not be stored because of the collision cause failures in returning values. Variant structures such as multi-hashing, cuckoo hashing, and d-left hashing have been studied, but none of the structures solve completely the collision problem. In this paper, we claim that a functional Bloom filter can replace a hash table. While the hash table requires to store each input key itself or the signature of each input key in addition to the return value, the functional Bloom filter can store the return value only, since different combinations of Bloom filter indexes can work as the signature of each input key. Performance evaluation results show that the functional Bloom filter is more efficient than hash-based data structures in storing more number of elements into a fixed size memory and hence in producing less failures.