Using Postgres FDW with a sharded PostgreSQL Database
summary
Shows how Postgres FDW links a 32-way sharded PostgreSQL setup (run on eight physical boxes) by creating foreign servers, per-server user mappings, and union all views that union per-shard foreign tables. Claims: WHERE on built‑in types pushes down, joins do not, queries hit shards serially (no parallel push), transactions span remote servers, remote defaults often don’t apply, and practical parallelism is achieved with external ETL/cron jobs.
AI-written summary · about
chapters (25)
- 0:00Intro
- 0:13Stephen Frost
- 0:48Do you read...
- 1:07What is an FOW?
- 2:36What is an FDW? (part 2)
- 3:32Basics of FDW connections
- 5:26Using postgres fow
- 6:41Createing User Mappings
- 7:24Creating Foreign Tables
- 8:36Remote Query Execution
- 10:38More on Query Execution
- 12:49Query costing with FOWS
- 15:53Dealing with 32 shards
- 17:16Sharding suggestions
- 19:21FDW Challenges
- 23:12Parallelizing
- 24:45Working through FDws
- 27:14View Example
- 27:56What's PG do?
- 29:12Playing with views
- 29:45Verbose
- 30:15Firing a remote procedure
- 31:31Managing foreign tables
- 33:55Improvements for FOWS
- 36:10More idle thoughts
related talks
52mSharding PostgreSQL using Foreign Data Wrappers
47mpg_shardman: Sharding in PostgreSQL
34mPostgreSQL sharding, state of the art
45mThe Future of Postgres Sharding
45mThe Future of Postgres Sharding
34mJoining Heterogeneous Databases is a reality, not a Myth
46mWriting a Foreign Data Wrapper
39mPostgreSQL: Joining Heterogeneous Databases is a reality, not a Myth