MergeTree

MergeTree is the default ClickHouse table engine — pretty much every analytical table you create uses it or a variant. The name is literal: data is written as sorted, immutable parts; a background process periodically merges small parts into bigger sorted parts, LSM-tree style. You specify ORDER BY (col1, col2, ...) — this determines the sort order on disk and the primary index (sparse, one mark per ~8000 rows by default). The family is where it gets interesting: ReplacingMergeTree — deduplicates rows with the same sort key during merges (late-arriving updates, or latest-wins patterns); SummingMergeTree — sums numeric columns on merge (pre-aggregated metrics); AggregatingMergeTree — stores partial aggregation states, merged later (backs most materialized views); CollapsingMergeTree — sign-based row cancellation for updates/deletes. ClickHouse is fundamentally "append fast, merge in background, query any part" — MergeTree is that idea made concrete.