Sharding for hyperscale and geographically distributed relational databases
summary
Describes Oracle Sharding implemented inside the database: horizontal, shared‑nothing shards with system-managed sharding and composite sharding options, direct and proxy routing, a shard catalog and shard director, and online rebalance. Claims a production deployment handling 1 million transactions per second over 2.5 petabytes with 1.6 ms reads and 2.5 ms writes. Announces raft replication, Kubernetes operator, Exadata optimization, online major-version upgrades, and an autonomous global hyperscale service.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 0:40What is Sharding
- 3:14Benefits
- 4:50Example
- 9:03How it works
- 12:24Components
- 13:20Partitioning data
- 15:17Raft replication
- 16:15Other replication options
- 16:52Multiregion configuration
- 17:29Golden Gate configuration
- 17:56Rebalance data
- 18:19Extra Data
- 18:41Kubernetes Operator
- 19:00Enterprise Manager Cloud Control
- 19:23Major Version Upgrades
- 19:47Backup and Recovery
- 20:11Migration
- 20:36Autonomous Global Database
- 22:14Additional Resources
- 22:32Outro
related talks
32mSharding for mission-critical, internet-scale apps
13mOracle Globally Distributed Database
48mHow to Use the new JDBC 4.3 Sharding API for Massive OLTP Scaling
46mDatabase Sharding the Right Way Easy, Reliable, and Open source
27mAutomatically shard & scale-out traditional databases on K8s
27mScaling for Intelligence: Distributed SQL Databases in the Age of AI
24mRethinking Database Architecture to Embrace Modern App Dev
42mPostgreSQL Distributed & Secure Database Ecosystem Building