MCF: Towards Window-Based Multiple Cuckoo Filter in Stream Computing

Abstract

In this paper, we present a new stream-oriented filter, named Multiple Cuckoo Filter (MCF), to support concise membership queries on multiple data streams. MCF is composed of a group of standard cuckoo filters, in which the membership query are decomposed into a set of single queries. MCF allows each cuckoo filter to be configured dynamically by changing the size of sliding window. It stores elements’ fingerprint, instead of elements themselves, here fingerprint is a bit string which is determined by a hash function. MCF can check whether a given item exists in multiple data streams simultaneously. It is proved that MCF outperforms better than traditional cuckoo filter on false positive in theory. Experiments demonstrate that the query time of MCF grows linearly with the growth of number of cuckoo filters, decreases gradually with the growth of sliding window number, and increases with the growth of total elements.