Data Pump - Oracle 10g/11g Data and Database Management Utilities - page 2

Oracle Data Pump is a facility available since Oracle 10g Release 1. It first appeared
back in 2003. It enables high speed data and metadata transfers. It is an efficient,
manageable, adaptive tool that can be used in more than one mode; namely, the
regular command line interface, the suspended mode, the network mode, and the
PL/SQL API mode. Besides the CLI interface, it is used by Enterprise Manager,
SQL*Loader (by means of the external data pump table driver), the PL/SQL API,
and other clients.
Data Pump is a productive tool designed to make the DBA's life easier. It can be
easily set to a suspended mode and brought back to work wherever it was stopped.
A session does not need an interactive connection to perform data management,
so it can leave an unattended job and it can be resumed any time. This tool doesn't
need to generate a file to transfer data in a database-to-database mode; it is the so
called network mode, which is very useful when a single load is performed. When
this data transfer mode is used, data does not have to be erased afterwards as there
is no intermediate file created to move the data. The network mode is similar to
the conventional named pipes which are used to perform data transfers on the fly;
however, this traditional approach is not available on all Operating Systems ( OSes )
(Windows does not support named pipes). If a task is launched, even if a degree
of parallelism hasn't been specified, it can be modified at run time, so resource
consumption can be increased or decreased at will.
Data Pump allows high speed data movement from one database to another. The
expdp command exports data and metadata to a set of OS files known as a dump
file set. Compared with the traditional export/import tool set, Data Pump allows a
DBA to easily clone accounts, move objects between tablespaces and change other
object features at load time without being required to generate an SQL script to have
the object modified, rebuilt and loaded. This kind of on-the-fly object redefinition is
known as the remap feature. Data Pump performance is significantly better than that
of the old export/import tools.
Data Pump architecture
Data Pump is a server side tool; even if it is remotely invoked, all the command
actions and file generation will take place on the host where the database resides,
and all directory objects refer to paths in the server. Oracle Data Pump requires
a Master Table which is created in the user's schema when a Data Pump session
is open. This table records the Data Pump's session status and if the job has to be
stopped (either on purpose or due to an unexpected failure), the Data Pump knows
where it was when it is brought back to work. This table is automatically purged
once the job is finished. The master table will match the job name given, by means of
the command line parameter job_name , or Oracle can choose to generate a name for
it, in case this parameter hasn't been defined.