Dragonfly rethinks Redis with sharded multi-threading to scale across modern multi-core servers
techNmak · x · 2026-09-03
Dragonfly is a rewrite of Redis built around modern multi-core servers rather than a single main thread. It partitions the keyspace into shards owned by different threads, letting the in-memory datastore actually use available CPU, memory and I/O.
Notable internals:
- A transaction framework inspired by the VLL lock-manager research enables atomic multi-key operations across shards without mutexes or spinlocks
- The core hash table draws on Dash research on scalable hashing
- Fork-less snapshotting replaces the traditional fork-based persistence approach
- Redis and Memcached compatibility is preserved for easy migration
More from Infra
- KV cache, not parameter count, may be the real bottleneck for long-context local models — jonejy · 2026-09-03
- Developers warn a new Cloudflare feature could hurt your SEO — turn it off — gaganghotra_ · 2026-09-03
- If Amazon Trainium is any good, why weren't they at Hot Chips? — firstadopter · 2026-09-03
- Zeiss exec: China is about 15 years behind in cutting-edge chipmaking tools — broodsugar · 2026-09-03
- Engineer tried 70B prefill + 7B decode back in 2023: combining spec dec with disagg was useless — suchenzang · 2026-09-03
- Blogger argues Zeiss's mirror-coating know-how is only ~30 years deep, not magic — teortaxesTex · 2026-09-03