Scaling for Intelligence: Distributed SQL Databases in the Age of AI
summary
Claims concrete outcomes and tools: uses Oracle Globally Distributed Database and distributed SQL patterns like sharding, raft quorum replication, and vector indexing; demonstrates an active‑active, strongly consistent system with automatic 3 second failover with zero data loss; shows a hyperscale deployment of 104 servers 5,000 CPU cores 77 TB memory and vector index scaling from 250M to 900M vectors at steady latency; reports that fine‑tuning smaller open models on private data outperforms larger public models.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:09Distributed SQL for AI
- 5:00Key properties
- 9:14Performance optimization
- 13:05Network resilience
- 15:12Global deployment
- 17:20SQL vs NoSQL
- 19:06Vector search scaling
- 21:57Model fine-tuning
- 24:47Evaluating distributed SQL
- 26:12Q&A
related talks
29mModern Database Architecture: From SQL to NoSQL & Distributed SQL
48mOracle Database Directions: Innovate, Scale, and Sleep Easier
23mSharding for hyperscale and geographically distributed relational databases
19mBuild Global-Scale Apps Using Globally Distributed Autonomous Databases
26mNewSQL: concepts, realities and when you should embrace it
18mData Superpowers with Oracle Database 23ai
24mFuture of Data and AI
11mDatabase Evolution: Partitioning and Indexing