A Smart Cuckoo Filter with Disjoint-set to Detect the Endless Loop

Cuckoo filter is a new type of approximate membership query data structure that stores the fingerprints of set elements. Because of its higher performance than bloom filter, cuckoo filter has been developed in recent years and gradually replaced the bloom filter in many fields. However, the endless loop problem in cuckoo hashing makes element insertion inefficient and unstable, which hinders cuckoo filter development. To solve this problem, we propose a new variant of cuckoo filter called smart cuckoo filter. Smart cuckoo filter uses a variable bucket structure to save the element insertion information at each position. It keeps buckets aligned to memory, thereby reducing the space waste and inefficiency of data access. Smart cuckoo filter uses the disjoint-set to determine whether the element will fall into the endless loop before the element is inserted. Also, the insertion is friendly to the concurrency of the entire filter. The experiments demonstrate that the element insertion speed of smart cuckoo filter in the endless loops is better than that of the traditional cuckoo filter, which can reach 2.2x in some cases.