ClickHouse excels at querying Parquet, a file format critical to Lakehouse architectures like Iceberg and Delta Lake. The engine directly queries files without ingestion and utilizes extensive parallelism and I/O reduction techniques. Current performance shows ClickHouse outpacing even popular native formats from other systems. A new native Parquet reader in development promises further improvements, including dictionary filtering and enhanced parallelism.
Table of contents
A Lakehouse-ready engine, by accident and by designInside the engine: How ClickHouse queries ParquetBenchmarking Parquet query performanceWrapping upSort: