Data Pump

Data Pump is the new high-speed infrastructure for data and metadata movement in Oracle Database 10g. The Data Pump commands are similar to the traditional export and import commands, but they are separate products. Data Pump provides a remarkable performance improvement over the original export and import utilities. It also provides faster data load and unload capability to existing tables. Using Data Pump, platform-independent flat files can be moved between multiple servers. You can use the new network mode to transfer data using database links.
Export and import operations in Data Pump can detach from a long-running job and reattach to it later with out affecting the job. You can also remap data during export and import processes. The names of data files, schema names, and tablespaces from the source can be altered to different names on the target system. It also supports fine-grained object selection using the EXCLUDE, INCLUDE, and CONTENT parameters.
Data Pump Export (dpexp) is the utility for unloading data and metadata from the source database to a set of operating system files (dump file sets). Data Pump Import (dpimp) is used to load data and metadata stored in these export dump file sets to a target database.
The advantages of using Data Pump utilities are as follows.
You can detach from a long-running job and reattach to it later without affecting the job. The DBA can monitor jobs from multiple locations, stop the jobs, and restart them later from where they were left.
Data Pump supports fine-grained object selection using the EXCLUDE, INCLUDE, and CONTENT parameters. This will help in exporting and importing a subset of data from a large database to development databases or from data warehouses to datamarts, and so on.
You can control the number of threads working for the Data Pump job and control the speed (only in the Enterprise version of the database).
You can remap data during the export and import processes.
In Oracle Database 10g Release 2, a default DATA_PUMP_DIR directory object and additional DBMS_DATAPUMP API calls have been added, along with provisions for compression of metadata in dump files, and the capability to control the dump file size with the FILESIZE parameter.