Adaptable Query Accuracy at Scale

Counting things at scale is deceptively hard. I showed how probabilistic data structures — Bloom Filters in particular — let distributed systems trade a small, configurable margin of error for orders-of-magnitude reductions in memory usage.

The talk walked through how platforms like Redis handle counting operations, why exact accuracy becomes prohibitively expensive at scale, and how to tune the accuracy-performance trade-off for real workloads.