Performance

Read your query plan as a graph

Altinity engineering·June 2026·4 min read

EXPLAIN PIPELINE tells you how ClickHouse will actually run a query. It just tells you in a wall of text.

ClickHouse has one of the better EXPLAIN families around. EXPLAIN PIPELINE in particular shows the physical processors — the things that actually read, transform, aggregate and merge your data — and how they fan out across threads. The catch is the output: an indented text tree that's precise and nearly unreadable past a dozen lines.

So the browser draws it. Click Explain on any query and you get five tabs — Explain, Indexes, Projections, Pipeline, Estimate — and the Pipeline tab lays the processor graph out as boxes and arrows. Pan it, zoom it, pop it fullscreen.

EXPLAIN PIPELINE rendered as a top-to-bottom processor DAG for a grouped, sorted query. — `EXPLAIN PIPELINE` for a `GROUP BY … ORDER BY`, drawn as a processor DAG.

Once it's a picture, the shape tells you things the text buries. Read the graph above top to bottom:

MergeTreeSelect × 4 — four parallel reading lanes. That × 4 is your parallelism: how many threads pull parts at once.
AggregatingTransform × 4 — the GROUP BY runs per-lane, in parallel, after a couple of expression steps.
Resize — the funnel. Parallel lanes collapse toward fewer. This is where parallel work becomes serial, and it's usually the first place to look when a query is slower than it should be.
PartialSorting → MergeSorting → MergingSortedTransform — the ORDER BY: sort each lane, then merge the sorted streams.

You start to read these the way you read a flame graph. Wide and parallel up top, a Resize in the middle, a merge at the bottom — that's a healthy aggregate-and-sort. A pipeline that stays one box wide the whole way down is single-threaded, and now you can see that instead of guessing at it.

It doesn't touch your query

A small thing I appreciate: hitting Explain doesn't rewrite your editor. The app wraps your statement in EXPLAIN behind the scenes and shows the result in the results pane — your SQL stays exactly as you typed it. The Indexes and Projections tabs run EXPLAIN indexes = 1 / projections = 1, which answer the question you usually have when a query scans more than it should: are my skip indexes and projections actually being used?

It's the kind of thing that's obvious the first time and slightly annoying to live without after. Paste a real GROUP BY into the live demo and hit Explain → Pipeline.