How ClickHouse became fast at joins

T1Covered by 1 source · first reported by Databricks / ClickHouse / DuckDB / Supabase / Neon / PlanetScale at 03 Jun, 00:00 UTC

Over two years of focused join engineering, ClickHouse became 26× faster on the TPC-H SF100 join-heavy workload. Here’s how parallel hash joins, runtime filters, lazy column replication, and smarter join planning got us there.