Operating Query Engines

Run query engines in production. EXPLAIN literacy mapped across Trino/ClickHouse/Druid/DuckDB to one mental model, a methodology for profiling a slow query, the OOM-vs-spill-vs-fail memory story, concurrency and admission control (resource groups, quotas, laning), multi-tenancy and isolation, caching layers and their staleness traps, autoscaling stateless vs stateful engines, capacity planning for p99 concurrency, per-query cost attribution, benchmarking honestly and designing your own, upgrades that flip query plans, and five recurring incident patterns. Ends with running an engine in production.