GoKawiilBluesky engineering details one-hash bloom filter design for Scylla lookups
Bluesky's engineering team published a technical writeup explaining how it uses bloom filters, including a one-hash-function variant, to avoid unnecessary database queries. The post describes how the company checks in-memory bloom filters before querying its Scylla database for block, mute and label lookups, citing roughly 1% false-positive rates that let 99% of such checks skip the database entirely.
GoKawiil's interpretation of the reporting above, not reported fact.
The approach illustrates how probabilistic data structures can let a large social platform cut database load at massive scale, which the company says has been important to operating the site cost-effectively. Because the writeup is a technical explainer rather than a product announcement, its significance is mainly for engineers building similarly high-throughput systems rather than for end users.
- Bloom filters let Bluesky skip millions of database queries per second for checks like blocks and mutes.
- Reported hit rates show about a 1% false-positive rate, meaning 99% of lookups avoid hitting the Scylla database.
- The post explores a one-hash-function bloom filter variant as an alternative to traditional multi-hash designs.
Source: calabro.io, 2026-10-19
Published there as: “One Hash Bloom Filters”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.