TransLattice Elastic Database Architecture Deep Dive
summary
Claims a production geographically distributed RDBMS built on PostgreSQL and Postgres-R, shipping standalone Q2 2012 after work since 2007. Architecture uses a global consensus protocol (generalized fast Paxos), MVCC, commit order IDs, encrypted/compressed networking, and shard-level stats with policy-driven placement and three-copy redundancy across failure domains. A six-node geographically dispersed cluster stands up in about an hour; distributed query processor and background workers run subqueries on remote nodes. Performance shows overhead for geo distribution but often lower latency for local users.
AI-written summary · about
chapters (11)
- 0:00Database challenges and TransLattice
- 6:48TransLattice Elastic Database overview
- 10:43Ted cluster deployment
- 17:11Global consensus and Paxos
- 21:53Data placement and redundancy
- 27:33PostgreSQL modifications for distributed ops
- 33:19User experience and query execution
- 40:54Distributed transaction management
- 43:48Sharding and partitioning
- 50:04Performance and use cases
- 54:33Paxos improvements and future plans
related talks
36mGeographically Distributed Prallel Databases - A Reality!
38mPostgreSQL as a distributed computing platform
42mPostgreSQL Distributed: Architectures & Best practices
51mEmbrace, Extend, Evolve - Making PostgreSQL Work for a Cloud Native World
49mPostgreSQL’s Journey to Distributed Databases
42mPostgreSQL Distributed & Secure Database Ecosystem Building
13mBuilding reliable infrastructure with Postgres
50mSupersized Postgres: Postgres-XL for OLTP and Big Data Analytics