What Is Parquet?
Apache Parquet is an open file format designed to store large datasets efficiently for analytics. It is column-oriented, meaning it stores the values of each column together rather than storing each row as a block. That single design choice makes a large difference for analytical queries, which typically read a few columns across many rows, and it is why Parquet has become a default storage format for data lakes and lakehouses. Despite the name it shares with flooring, in data it refers only to this format.
Why Columnar Storage Matters
Most analytics reads a handful of columns from a very large table: the total of a sales column, the average of a price column. A row-based format has to read every row in full to get those columns, which is wasteful. A columnar format like Parquet reads only the columns a query needs, skipping the rest. Because values in a column are similar to each other, they also compress extremely well, so Parquet files are smaller and faster to scan than row-based equivalents.
Key Characteristics
Parquet is known for a few things in particular:
- Columnar layout, so queries read only the columns they need.
- Strong compression, because similar values sit together.
- A schema stored with the data, so each file describes its own structure.
- Broad support, readable and writable by most data engines and platforms.
Together these make Parquet efficient and portable, which is much of its appeal.
Parquet and the Lakehouse
Parquet is one of the building blocks of the lakehouse. Data is stored as Parquet files on inexpensive cloud storage, and an open table format such as Apache Iceberg or Delta Lake organizes those files into reliable tables with transactions and schema evolution. In that stack, Parquet is the layer that holds the actual data efficiently, while the table format adds the database-like guarantees on top.
Parquet and Processing Engines
Because Parquet is open and widely supported, the same files can be read by many engines, Apache Spark, SQL query engines, and BI tools among them, without converting the data first. That portability is a major reason organizations store analytical data in Parquet: it keeps the data usable across tools and avoids locking it into one vendor’s proprietary format.
Open Storage in an ERP Data Foundation
Storing modeled ERP data in an open format like Parquet keeps the foundation portable and efficient. The data can be queried quickly and read by whatever tools the business chooses, now and later. QuickLaunch builds governed foundations on open lakehouse storage for JD Edwards, Vista, NetSuite, and OneStream, so the data is held efficiently and stays open rather than locked to one platform.
Frequently Asked Questions
What is Parquet?
An open, column-oriented file format for storing large datasets efficiently. By storing each column together, it lets analytical queries read only the columns they need and compress the data well, which makes it fast for analytics.
Why is Parquet good for analytics?
Because analytics usually reads a few columns across many rows. Parquet’s columnar layout reads only the needed columns and skips the rest, and similar values in a column compress strongly, so files are smaller and faster to scan than row-based formats.
What is the difference between Parquet and Iceberg?
Parquet is a file format that stores the data. Apache Iceberg is a table format that organizes many Parquet files into a reliable table with transactions, schema evolution, and time travel. They work together: Iceberg tables are usually made of Parquet files.
About the Author
