Spanner - Google's Distributed Database
summary
Spanner claims a planet-scale, transactional database that partitions and replicates across thousands of servers and still gives row/column models and SQL. It uses GPS and atomic clocks to assign globally meaningful timestamps, enabling lock-free strong reads and global snapshots consistent with external events. The system implements Paxos-based replication, per-table sharding (tablets), and client-directed reads; engineers report it works at production scale with bounded clock uncertainty and acceptable extra hardware.
AI-written summary · about
chapters (8)
- 0:00Spanner introduction
- 2:31Spanner architecture overview
- 9:54Spanner transactions and concurrency
- 15:05Consistent snapshots in distributed systems
- 22:42TrueTime in Spanner
- 29:00TrueTime implementation with GPS and atomic clocks
- 30:03Spanner development lessons
- 33:39Q&A on clock accuracy and testing
related talks
49mSpanner - fully managed horizontally scalable relational DB with ACID transactions
47mGoogle Cloud Spanner: Global consistency at scale
33mGoogle Cloud Spanner: Global consistency at scale
37mGoogle Cloud Spanner: Going global with relational database
01h 08mKeynote: Building Spanner
22mOrder And Consistency At Scale
01h 13mExtending PostgreSQL to Google Spanner architecture
49mPostgreSQL’s Journey to Distributed Databases