Writing Parquet Files
The Parquet data format (https://parquet.apache.org/) is binary, columnar storage that can be used by different tools, including Hadoop and Spark. It was built to support compression, to enable higher performance and storage use. Its column-oriented design helps with data selection for performance, as only the data in the required columns are retrieved, instead of searching for the data and discarding values in rows that are not required, reducing the retrieval time for big data scenarios, where the data is distributed and on disk. Parquet files can also be read and written by external applications, with a C++ library, and even directly from pandas.
The Parquet library is currently being developed with the Arrow project (https://arrow.apache.org/).
When considering more complex queries in Spark, storing the data in Parquet format can increase performance, especially when the queries need to search a massive dataset. Compression helps to decrease the data volume that needs...