Loading and Unloading Data
A newer version of this documentation is available. Click here to view the most up-to-date release of the Greenplum 4.x documentation.
Loading and Unloading Data
The topics in this section describe methods for loading and writing data into and out of a Greenplum Database, and how to format data files.
Greenplum Database supports high-performance parallel data loading and unloading, and for smaller amounts of data, single file, non-parallel data import and export.
- The COPY SQL command transfers data between an external text file on the master host and a Greenplum Database table.
- External tables allow you to query data outside of the database directly and in parallel using SQL commands such as SELECT, JOIN, or SORT EXTERNAL TABLE DATA, and you can create views for external tables. External tables are often used to load external data into a regular database table using a command such as CREATE TABLE table AS SELECT * FROM ext_table.
- External web tables provide access to dynamic data. They can be backed with data from URLs accessed using the HTTP protocol or by the output of an OS script running on one or more segments.
- The gpfdist utility is the Greenplum Database parallel file distribution program. It is an HTTP server that is used with external tables to allow Greenplum Database segments to load external data in parallel, from multiple file systems. You can run multiple instances of gpfdist on different hosts and network interfaces and access them in parallel.
- The gpload utility automates the steps of a load task using a YAML-formatted control file.
- The GemFire-Greenplum Connector allows the transfer of data between a Pivotal GemFire region and a Greenplum Database table. Pivotal Gemfire is an in-memory data management system that provides reliable asynchronous event notifications and guaranteed message delivery. For information about using GemFire-Greenplum Connector, see http://ggc.docs.pivotal.io/. For information about Pivotal Gemfire, see http://gemfire.docs.pivotal.io/.
In the simplest case, the COPY SQL command loads data into a table from a text file that is accessible to the Greenplum Database master instance. This requires no setup and provides good performance for smaller amounts of data. With the COPY command, the data copied into or out of the database passes between a single file on the master host and the database. This limits the total size of the dataset to the capacity of the file system where the external file resides and limits the data transfer to a single file write stream.
More efficient data loading options for large datasets take advantage of the Greenplum Database MPP architecture, using the Greenplum Database segments to load data in parallel. These methods allow data to load simultaneously from multiple file systems, through multiple NICs, on multiple hosts, achieving very high data transfer rates. External tables allow you to access external files from within the database as if they are regular database tables. When used with gpfdist, the Greenplum Database parallel file distribution program, external tables provide full parallelism by using the resources of all Greenplum Database segments to load or unload data.
Greenplum Database leverages the parallel architecture of the Hadoop Distributed File System to access files on that system.