← All comparisons

Parquet vs ORC

A side-by-side feature comparison.

Summary

Parquet and ORC are the two columnar file formats that dominate analytics. They have nearly identical capabilities — column-grouped storage, multiple compressors, predicate pushdown, schema evolution — but Parquet won mindshare outside Hive and is the safer default in 2026.

Bottom line: Parquet for general analytics. ORC inside an established Hive warehouse.

Feature comparison

Feature Apache Parquet Apache ORC
Origin Twitter / Cloudera Hortonworks
Compression Snappy / GZIP / Zstd / LZ4 Zstd / Snappy / Zlib / LZ4
Predicate pushdown Yes Yes
Index granularity Row group Stripe + index
Adoption Industry-wide Hive ecosystem
Schema evolution Yes Yes

About Apache Parquet

Parquet is the dominant columnar data file format for analytics. Each file groups rows into row-groups and stores each column as its own compressed page, allowing query engines to read only the columns they need. DuckDB, Spark, BigQuery, Snowflake and pandas all read and write Parquet natively.

Read the full Apache Parquet guide →

About Apache ORC

ORC (Optimized Row Columnar) is a Hadoop-ecosystem columnar format from Hortonworks, designed for Hive workloads. Like Parquet, it stripes data column-by-column with rich indexing and statistics, but with a different row-group layout that often performs better on Hive itself. Outside the Hive world, Parquet is more common.

Read the full Apache ORC guide →

More comparisons