In the Blaze engine, map-side aggregation is analogous to the aggregation done in the map phase in a MapReduce job that runs on the Hive engine. Source data is aggregated based on the group-by port set in an Aggregator transformation. The aggregated data then moves to the data shuffle stage for the second level of aggregation.

If you specify a unique key as the group by port, disable the map-side aggregation in a mapping that runs on the Blaze engine.

In the Developer tool, set the following run-time property in the Run-time tab to disable map-side aggregation for the mapping: