User Guide

11.0
- 10.0

Back Next

Data Unload Methods

Fast Clone supports conventional path and direct path unload methods.

The direct path unload method reads data directly from Oracle data files and requires the Oracle user to have special user permissions, such as SELECT FROM DICTIONARY. This method provides high-performance unloads of data without increasing overhead on the source system. With this method, you can use the DataStreamer component to additionally enhance performance. By default, Fast Clone does not lock the source tables to unload data. However, you can change this behavior to perform a consistent read of the source data.

Fast Clone cannot read data from Oracle data files that are on a standalone system. Fast Clone must connect to a running Oracle instance to retrieve metadata for unload processing.

The conventional path unload method uses the OCI to retrieve metadata and data from the Oracle source. This method usually provides an acceptable and efficient level of performance but is slower than the direct path unload method. With the conventional path unload method, you can run Fast Clone on a system that is remote from the source database server without any special permissions.

The following table describes the differences between the direct path unload and conventional path unload methods:

Feature	Direct Path Unload	Conventional Path Unload
Reads data directly from the Oracle source files	Yes	No
Performs high performance unloads	Yes. Provides very high-performance unloads, up to 20 times faster than the native database exports.	Yes. When Fast Clone unloads data locally, performance is good but CPU consumption can be twice that of the direct path unload method. When Fast Clone unloads data remotely, performance might be slower than that of unload on a local system.
Supports remote execution of Fast Clone	Yes, only if the Fast Clone Server is used.	Yes
Supports SQL queries that perform joins between tables	No	Yes
Works with DataStreamer	Yes	No
Unloads data from Oracle Exadata Database Machine	No	Yes. Fast Clone must use one thread per table to unload data from Oracle Exadata.
Unloads data from views and cluster tables	No	Yes
Performs parallel processing of data unloads using multiple threads	Yes	Yes. For partitioned tables, Fast Clone uses one thread per table partition or subpartition. For tables that are not partitioned, Fast Clone distributes work across threads based on the Oracle ROWID value, provided that unload processing is local.
Integrates with Informatica Data Replication	Yes	Yes

Rename Saved Search

Table of Contents

User Guide

User Guide

Data Unload Methods

Data Unload Methods