Description

Use Spark Aggregator instead of the current implementation.
The de-duplication functionality could be added to the aggregator.

Related issues:
pth_10#581, pth_10#575