Optimizing Cuckoo Filter for high burst tolerance,low latency, and high throughput

TL;DR

This paper introduces an optimized cuckoo filter designed for distributed data stores, achieving high burst tolerance, low latency, and high throughput by improving search efficiency and reducing false positives.

Contribution

The paper presents a modified cuckoo filter implementation that enhances search performance and false positive rates for high workload distributed databases.

Findings

Improved search times in distributed data stores

Reduced false positive rates compared to traditional filters

Enhanced support for delete operations

Abstract

In this paper, we present an implementation of a cuckoo filter for membership testing, optimized for distributed data stores operating in high workloads. In large databases, querying becomes inefficient using traditional search methods. To achieve optimal performance it is necessary to use probabilistic data structures to test the membership of a given key, at the cost of getting false positives while querying data. The widely used bloom filters can be used for this, but they have limitations like no support for deletes. To improve upon this we use a modified version of the cuckoo filter that gives better amortized times for search, with less false positives.