PostgreSQL extension (pronunciation unclear in the talk) that merges several extensions into one, most notably columnar storage, distributed tables and parallel querying. Columnar storage is append-only but compresses typically ~10× and is much faster for analytical/aggregate queries; enabled simply with `USING columnar`, and columnar/row tables can be mixed in a single query (e.g. cold-partition columnar + newest-partition row). Distributed tables spread data across multiple cheaper Postgres servers via `create_distributed_table(table, column)` — you pick a distribution column (customer_id in a SaaS), use collocation to keep related tables together, and mark small shared tables as reference tables so they replicate to every server.