Features

The Systems section is organized by directory. This section is organized by capability — the cross-cutting features a user actually thinks about.

Pages

• Replication — ReplicatedMergeTree, the replication log, fetches, quorum, parallel replicas (read side).
• Distributed queries — Distributed tables, sharding, cluster discovery, ON CLUSTER, parallel replicas, hedged connections.
• Materialized views — incremental and refreshable MVs, WindowView, LiveView, target tables.
• Projections and TTL — projections, materialized indices, table/column/move TTLs.
• Async insert — server-side batching of small inserts.
• Text and vector search — full-text indexes, vector similarity index, related functions.
• Object storage tables — S3/Azure/HDFS, Iceberg, DeltaLake, Hudi, Hive.

Each feature page links back to the systems pages where the heavy code lives, so you can follow the rabbit hole as deep as you want.

How to find what implements a feature

The natural starting points:

• Engine-bounded features (replication, projections, lightweight delete) → src/Storages/MergeTree/.
• Read-time features (distributed queries, parallel replicas, hedged connections) → src/Interpreters/ClusterProxy/, src/Client/.
• DDL-driven features (RBAC, named collections, refreshable MVs) → src/Interpreters/ and src/Access/.
• Storage-substrate features (S3 cache, encrypted disk, Iceberg) → src/Disks/, src/Storages/ObjectStorage/.