Integration Services uses transformations to manipulate data during an ETL dataflow. Transformations can be used in [[articles:SSIS Dataflow Task|Data Flow Tasks]], between data source and destination components, or other transformations.

Sometimes the source data has large volumes, and during development, less rows would be better to keep the speed of the tests high. The Row Sampling transformation can be used for random sampling, to create a small but representative dataset. A fixed number of rows can be selected using Row Sampling, defined in the SamplingValue property - which can be overridden with SSIS Expressions.

If a percentage of the input rows should be selected by a transformation (instead of a fixed number), check the Percentage Sampling Transformation.

This transformation is a row (synchronous) transformation, and supports one input and two outputs: the sample output and the unselected output. It does not support an error output since data values on the outputs do not depend on any evaluation of expressions.

For more detailed information on this transformation, please refer to this MSDN article.
For the complete list of SSIS transformations, check this Wiki article.