What is Parquet

Definition

Apache Parquet is an open-source columnar storage file format designed for analytics, storing data by column with built-in compression and encoding so that query engines can read only the needed columns and skip irrelevant data efficiently at large scale.

« Back to Glossary Index

Stores data by column so analytical queries read only the fields they need
Achieves high compression through column-wise encoding of similar values
Supports predicate pushdown and data skipping to avoid scanning irrelevant blocks
Works across virtually every modern analytics engine as an open standard

Real World Example

A data lake stores its event tables as Parquet, so a Spark query filtering on date and summing one metric reads only the relevant column chunks, cutting both scan volume and cost compared with raw JSON.

FAQs

Why is Parquet preferred for analytics?

Its columnar layout, compression, and ability to read only needed columns make analytical queries far faster and cheaper than row formats.

How does Parquet differ from CSV?

Parquet is binary, columnar, compressed, and schema-aware, while CSV is plain text and row-oriented, making Parquet far more efficient at scale.

What engines read Parquet?

Spark, Trino, BigQuery, Snowflake, and most modern analytical engines support Parquet natively.