Bulk Operation

SPARK-9850 Allow shuffle readers to request data from just one mapper

SPARK-9850 Improve the way that we plan shuffled join

SPARK-9850 Join: Handling data skew

SPARK-9850 Join: Determine the number of reducers used by a shuffle join operator at runtime

SPARK-9850 Join: Determine the join strategy (broadcast join or shuffle join) at runtime

SPARK-9850 Aggregation: Determine the number of reducers at runtime

SPARK-9850 Introduce an ExchangeCoordinator to estimate the number of post-shuffle partitions.

SPARK-9850 Let reduce tasks fetch multiple map output partitions

SPARK-9850 Support submitting map stages individually in DAGScheduler