In addition to reading flat files and interfacing with access modules, the DataConnector operator also has the ability to read Hadoop files and tables. The following table briefly describes and compares the two interfaces which the DataConnector operator can use to move data from Hadoop files and tables to the data stream, and eventually the database.
Interface | Description |
---|---|
HDFS API | Provides access to Hadoop files via the Hadoop Distributed File System Application Programming Interface, or HDFS API. The HDFS is a POSIX compatible file system with some minor restrictions. It does not support updating files and it only supports writing files in truncate mode or append mode. The Hadoop Software is written in Java and the HDFS API is a Java JNI interface that exposes all the expected standard posix file system interfaces for reading and writing HDFS files directly by a C++ program. The Data Connector Producer and Consumer have been updated to directly access the HDFS file system using the HDFS API. All standard Data Connector file system features are supported. |
TDCH-TPT | Provides access to Hadoop files and tables via the Teradata Connector for Hadoop or TDCH. TDCH utilizes the MapReduce framework's distributed nature to transfer large amounts of data in parallel from Hadoop files and tables to the DataConnector operator. TDCH has the ability to read and write HDFS files, Hive tables, and Hcat tables stored in various Hadoop-specific formats. Because this interface relies on TDCH to read and write data, many of the traditional DataConnector attributes are unsupported when using the TDCH-TPT interface. |
For more information, see “Processing Hadoop Files and Tables” in Teradata® Parallel Transporter Reference, B035-2436.
- GZIP and ZIP files are not supported with Hadoop/HDFS.
- HDFS processing can be activated simply by adding the following attribute to a Data Connector Consumer or Producer: HadoopHost = 'default’