Mastering Apache Storm

CombinerAggregator

The CombinerAggregator first runs the partitionAggregate on each partition, then runs the global repartitioning operation to combine all the partitions of the same batch into a single partition, and then reruns the aggregator on the final partition to emit the desired output. The network transfer here is less compared to the other two aggregators. Hence, the overall performance of the CombinerAggregator is better than the Aggregator and ReduceAggregator.

The CombinerAggregator<T> interface contains the following methods:

• init(): This method runs on each input tuple to retrieve the fields' value from the tuple.
• combine(T val1, T val2): This method combines the values of the tuples. This method emits a single tuple with ...