Description
Currently, SparkSQL only support the hash-based aggregation, which may cause OOM if too many identical keys in the input tuples.
Attachments
Issue Links
- is blocked by
-
SPARK-2926 Add MR-style (merge-sort) SortShuffleReader for sort-based shuffle
- Resolved
- is duplicated by
-
SPARK-2873 Support disk spilling in Spark SQL aggregation
- Resolved
- links to