Github user gczsjdy commented on the issue: https://github.com/apache/spark/pull/19763 Actually, the time gap is O(number of mappers * shuffle partitions). In this case, number of mappers is not very large, while users are more likely to get slowed down when they run on a big data set.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org