Github user BryanCutler commented on the issue: https://github.com/apache/spark/pull/21546 Hey @HyukjinKwon , after going through the previous benchmarks, it seems out-of-order batches had more of an effect on performance that I thought with `toPandas`. The current revision of this PR (which buffers out of order batches in the driver JVM) has about a 1.06x - 1.09x speedup which seems a bit underwhelming after getting ~1.25x when sending out-of-order batches. I still want to try to verify the old numbers and will hopefully get to that tomorrow.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org