CPCF: A Flexible Chunking and Proactive Insertion Cuckoo Filter

Wendi Hua, Ping Xie, Xiao Qin, Rui Mao, Yan Ji, Hao Chen, Hua Yan, Jianbin Qin

Abstract:

Cuckoo filter is a data structure for approximate membership queries widely used in various data science fields. However, inefficient space usage and element insertion prevent cuckoo filters from completely replacing Bloom filters. We present CPCF, a new and efficient version of cuckoo filter, which improves space utilization and insertion speed without any sacrifice. CPCF employs flexible chunking to optimize space efficiency. It automatically adjusts chunk sizes to the number of elements while minimizing granularity. A proactive insertion strategy accelerates insertion with reduced moving hash conflict elements. CPCF also astutely detects hashing failure, enhancing insertion stability. Experiments show that CPCF conserves more space than the state-of-the-art cuckoo filter variant in most cases. Additionally, CPCF augments insertion throughput by 21%
101% under maximum load compared with other variants. The dynamic thresholds ensure accurate judgment of hashing failures at lower values. These optimizations render CPCF a versatile and high-performance approximate membership query filter.