Projects

Systems · 2025

Bloom and Cuckoo Filters

A benchmark comparing Bloom filters, counting Bloom filters, and cuckoo filters under skewed access patterns instead of uniform textbook traffic.

Python · probabilistic data structures

The question

I wanted to know how the structures behave when a small number of keys are accessed much more frequently than the rest.

How it works

I built a Python benchmarking harness that runs all three structures against the same Zipf-distributed workloads and measures memory use, lookup latency, and throughput.

Outcome

Under the workloads I tested, cuckoo filters produced lower lookup latency and higher throughput. The project was more about understanding the tradeoffs between the structures than simply reimplementing each one.