Azure Cosmos DB - Low Latency and High Availability at Planet Scale
summary
Azure Cosmos DB runs a multi-API, multi-region storage system that guarantees every write is durably flushed to at least three of four replicas before acknowledging, uses Paxos for leader election, and supports five tunable consistency levels. Partitioning drives horizontal scale (hierarchical partitions to reduce cross-partition queries), RU-based resource governance with provisioned/autoscale/serverless billing, and an API Gateway plus a direct TCP connection mode for SDKs. One customer operates at 100 million requests per second across petabytes in 40+ regions; topology changes return 410 to trigger SDK reconnection.
AI-written summary · about
chapters (11)
- 0:00Introduction and history
- 1:44Cosmos DB features
- 6:13High-level architecture
- 7:32Resource hierarchy
- 11:52Replicas and partitions
- 13:18Fault domains and zones
- 16:08Write request flow
- 25:01Read consistency and scaling
- 47:18Elasticity and APIs
- 51:02Gateway and load balancing
- 58:21SDK functionality
related talks
16mAzure Cosmos DB: A look at the Universe
01h 13mBuilding Planet Scale .NET Apps with Azure Cosmos DB
01h 03mAzure Cosmos DB - The Best NoSQL Database You're Probably Not Using (Yet)
02h 02mAzure Cosmos DB: Liftoff
45mGetting the Most from NoSQL in the Cloud with Azure Cosmos DB
03h 30mAmericas Live Stream
05mLessons learned from building a globally distributed database service...
50mAzure Cosmos DB - Technical Deep Dive
