Reusable Distributed Data Infrastructure with Dynomite
summary
Dynomite converts single-node stores into a peer‑to‑peer, master‑master layer that adds sharding, replication and multi‑DC failover to Redis or RocksDB, with client library Dino and orchestration via Dynomite Manager and Explorer. Netflix runs 70 clusters across ~1,000 nodes, typically nine replicas (three racks × three regions), handles millions of OPS, supports zero‑client‑downtime node termination via streaming warm‑up, and notes ongoing work on reconciliation and RocksDB tuning.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 0:43What is Dynomite
- 2:07Netflix’s need for Dynomite
- 3:34Dynomite architecture
- 6:25Peer replication & multi-DC
- 7:47Replication & token ranges
- 9:11Dino client features
- 11:18Dynomite Manager
- 12:41Switching storage engines
- 13:44Node warm-up & migration
- 14:26Dynomite Explorer
- 15:28Future roadmap
- 16:54Open source adoption
- 17:36Q&A
related talks
59mUsing Redis beyond caching
41mEverything you always wanted to know about highly available distributed databases
43mDYNAMO: the paper that changed the database world
30mData @Scale Seattle- Muthu Annamalai
20mHow to Build a High Performance App Using Cloud Foundry and Redis
24mHow Netflix enables
30mBuilding domain specific databases with a distributed commit log
49mData Platform Architecture, Evolution, and Philosophy