A columnar storage file format optimized for large-scale data processing and analytics workloads, widely used in big data and cloud environments.
Unlike row-based formats like CSV (which store all values for one record together), Parquet stores all values for one column together. This columnar layout makes it much faster to read only the columns needed for a query, and enables efficient compression because similar values are stored adjacent to each other.
Parquet is the preferred format for large datasets in analytics platforms like Apache Spark, AWS Athena, and Google BigQuery. It is increasingly used in open data publishing for large datasets where performance matters.
Example: A national dataset of 50 million property transactions is published in both CSV and Parquet formats. A data scientist who needs only the sale price and date columns can read the Parquet file and retrieve just those two columns without loading the entire dataset into memory — a query that takes seconds in Parquet might take minutes with CSV.