Fast Clone uses a named pipe that has fixed size of approximately 6 MB for each source table from which data is unloaded. By default, the pipe name is based on the name of the source table or table partition. After Fast Clone writes a chunk of data to the pipe, the target load utility reads the data from the pipe and loads it to the target. Fast Clone then writes the next chunk of data to the pipe. After the load utility loads the last chunk of data from the pipe, Fast Clone closes the pipe. If the load utility does not read the data from the pipe, Fast Clone unload processing remains pending.

To load data from named pipes to targets, you must generate the load script before you run the unload job. By default, Fast Clone generates the load script after it completes unloading data from the source to output files. However, to stream data through pipes to the target database, you must generate the load script, start unload processing, and then run the load script while unload processing is in progress. To generate the load script and unload data to pipes, you must run the Fast Clone twice:

Before you begin unload processing, either manually create a pipe for each source table in the output directory or configure Fast Clone to create the pipes. If you configure Fast Clone to create the pipes, monitor pipe creation in the output directory so that you do not start the load jobs until after the pipes are created. Alternatively, you can create a script to delay load processing until after Fast Clone creates all of the pipes.