Data format options for inputs and outputs in AWS Glue - AWS Glue

Data format options for inputs and outputs in AWS Glue

These pages offer information about feature support and configuration parameters for data formats supported by AWS Glue. See the following for a description of the usage and applicablity of this information.

Feature support across data formats in AWS Glue

Each data format may support different AWS Glue features. The following common features may or may not be supported based on your format type. Refer to the documentation for your data format to understand how to leverage our features to meet your requirements.

Read AWS Glue can recognize and interpret this data format without additional resources, such as connectors.
Write AWS Glue can write data in this format without additional resources. You can include third-party libraries in your job and use standard Apache Spark functions to write data, as you would in other Spark environments. For more information about including libraries, see Using Python libraries with AWS Glue.
Streaming read AWS Glue can recognize and interpret this data format from an Apache Kafka, Amazon Managed Streaming for Apache Kafka or Amazon Kinesis message stream. We expect streams to present data in a consistent format, so they are read in as DataFrames.
Group small files AWS Glue can group files together to batch work sent to each node when performing AWS Glue transforms. This can significantly improve performance for workloads involving large amounts of small files. For more information, see Reading input files in larger groups.
Job bookmarks AWS Glue can track the progress of transforms performing the same work on the same dataset across job runs with job bookmarks. This can improve performance for workloads involving datasets where work only needs to be done on new data since the last job run. For more information, see Tracking processed data using job bookmarks.

Parameters used to interact with data formats in AWS Glue

Certain AWS Glue connection types support multiple format types, requiring you to specify information about your data format with a format_options object when using methods like GlueContext.write_dynamic_frame.from_options.

Some connection types do not require format_options. For example, in normal use, a JDBC connection to a relational database retrieves data in a consistent, tabular data format. Therefore, reading from a JDBC connection would not require format_options.

Some methods to read and write data in glue do not require format_options. For example, using GlueContext.create_dynamic_frame.from_catalog with AWS Glue crawlers. Crawlers determine the shape of your data. When using crawlers, a AWS Glue classifier will examine your data to make smart decisions about how to represent your data format. It will then store a representation of your data in the AWS Glue Data Catalog, which can be used within a AWS Glue ETL script to retrieve your data with the GlueContext.create_dynamic_frame.from_catalog method. Crawlers remove the need to manually specify information about your data format.

For jobs that access AWS Lake Formation governed tables, AWS Glue supports reading and writing all formats supported by Lake Formation governed tables. For the current list of supported formats for AWS Lake Formation governed tables, see Notes and Restrictions for Governed Tables in the AWS Lake Formation Developer Guide.

Note

For writing Apache Parquet, AWS Glue ETL only supports writing to a governed table by specifying an option for a custom Parquet writer type optimized for Dynamic Frames. When writing to a governed table with the parquet format, you should add the key useGlueParquetWriter with a value of true in the table parameters.