Parquet

Apache Parquet is an open source, column-oriented data file format built for efficient storage and retrieval of large analytical datasets. Instead of writing rows one after another, Parquet stores values by column, which lets query engines read only the columns they need and apply aggressive, type-aware compression and encoding. It has become a default interchange format across the big-data and analytics ecosystem.

• Columnar layout - Values are grouped by column so scans touch only the fields a query cares about, not entire rows.
• Efficient compression and encoding - Per-column schemes like dictionary and run-length encoding shrink storage and speed up reads.
• Rich, self-describing schema - Each file carries its own typed schema and nested-structure support, so consumers do not need an external contract to read it.
• Broad engine support - Spark, DuckDB, Arrow, and most cloud data warehouses read and write it natively, making it a portable lingua franca for data at rest.

In API operations Parquet tends to show up on the analytics and bulk-data side rather than the request/response path — it is what a data or dataset API hands back for large exports, what feeds pipelines behind an API, and what pairs with partitioning to keep those datasets fast to query. As more API traffic becomes machine- and agent-driven, formats like Parquet matter because they let systems move columnar data at scale without the overhead of row-oriented text formats. It is less an API standard than a storage and interchange substrate the data-heavy corners of the API economy quietly depend on.