Extending PostgreSQL to Google Spanner architecture
summary
YugabyteDB claims to deliver full PostgreSQL compatibility by reusing PostgreSQL upper half while replacing storage with a Spanner-like DocDB that shards tables into tablets and replicates each tablet with Raft replication and leader leases. Distributed transactions use a two‑phase commit with hybrid logical clocks and a transaction status table; leader election and failover target ~3s. RocksDB is modified for key→document, global block cache and batching; system catalogs are co‑located; automatic tablet splitting and multi‑region zero‑RPO operation are supported.
AI-written summary · about
chapters (17)
- 0:00Start
- 4:33design + Aurora-like Compatibility
- 6:36ode PostgreSQL
- 16:31Sharding Examples
- 19:00Membership Changes
- 24:10Leases to the Rescue!
- 25:50onic Clocks
- 30:46Encoding of a Document
- 33:40erformance Enhancements
- 39:08one Deployment
- 42:17ng Zone Outage
- 43:52tic rebalancing
- 48:42centralized Architecture
- 54:21ted (aka Multi-Shard) Transactions a
- 54:45Distributed Transactions need?
- 56:34Logical Clock (HLC)
- 59:35PostgreSQL Architecture
related talks
30mTechnical Comparison of YugabyteDB with Oracle and PostgreSQL
27mIntroduction to YugabyteDB: Design and Architecture
26mIntroducing YugabyteDB for PostgreSQL Users
49mPostgreSQL’s Journey to Distributed Databases
01h 27mYugabyteDB YSQL Development
25mYugabyteDB Technical Overview
47mAnatomy of a Distributed SQL Database - YugabyteDB
27mIntroduction to YugabyteDB – Design and Architecture