Sharding for mission-critical, internet-scale apps
summary
Oracle Sharding shows a production site with 2.5 petabytes of data, a 22 billion‑row table, 30 billion API calls/day, end‑to‑end reads at 1.6 ms and writes at 2.5 ms, 180 TB/hour of redo, and a 52‑instance deployment totaling 2,704 cores and 38,740 GB RAM. It shards by chunks using composite sharding (geo list + per‑region hashing), routes via enhanced drivers and a shard director, offers a Sharding Advisor, Kubernetes/Terraform automation, centralized backup with GDSCTL, online chunk rebalancing, shard‑replacement upgrades and persistent‑memory support.
AI-written summary · about
chapters (18)
- 0:00Intro
- 0:31Oracle Bluekal scales to 1 Million transactions per second with Oracle Database Sharding
- 5:37Benefits of Sharding
- 8:02Key Customer Use Cases
- 10:02Oracle Sharding - Better deal for Customers
- 12:11Schema - Sharded and Duplicated Tables
- 15:16Concept: Chunk
- 16:28Application Request Flow
- 18:47HA/DR Configurations
- 19:28Sharding Methods
- 23:09Online Addition and Rebalancing of Shards
- 25:07Sharding with Exadata
- 25:42Automatic Identification of Sharding Key
- 26:33Deployment Automation
- 27:58Centralized Backup Recovery
- 29:05Upgrade using Shard Replacement
- 29:42Support for Databases in Persistent Memory
- 31:25Summary
related talks
23mSharding for hyperscale and geographically distributed relational databases
13mOracle Globally Distributed Database
48mHow to Use the new JDBC 4.3 Sharding API for Massive OLTP Scaling
52mTo Shard or Not to Shard
46mDatabase Sharding the Right Way Easy, Reliable, and Open source
37mSharding: Growing Systems from Node-scale to Planet-scale
52mDatabase sharding at the load balancer level
27mAutomatically shard & scale-out traditional databases on K8s