Query engine

A query engine is the component that takes a query, most often written in SQL, and turns it into an execution plan that actually retrieves and computes the result — parsing the query, deciding an efficient order of operations, and then running that plan against underlying storage. Some query engines own their storage, like a traditional database; others, sometimes called federated or lakehouse query engines, run entirely separately from storage and read data wherever it lives.

A key design choice is whether a query engine is coupled to a specific storage layer or decoupled from it: decoupled engines such as Trino or open-source Spark SQL can query files in a data lake directly, using an open table format for transactional guarantees, without owning the data themselves, which is what enables data virtualization across multiple sources. The engine's optimizer, the part responsible for query optimization, is usually the biggest factor separating fast and slow implementations.

Choosing a query engine matters because it determines what data can be queried together, how much data movement a query requires, and how well it scales — most production engines rely on massively parallel processing to spread work across nodes. A common confusion is treating "query engine" and "data warehouse" as synonyms; a warehouse typically bundles a query engine with its own managed storage, while a standalone query engine can be pointed at storage it does not own.