Hi Spark devs, Kyle identified a deficiency in Spark where generating iterators are unrolled into memory and then flushed to disk rather than sent straight to disk when possible.
He's had a patch sitting ready for code review for quite some time now (100 days) but no response. Is this something that an admin would be able to review? I for one would find this quite valuable. Thanks! Andrew https://spark-project.atlassian.net/browse/SPARK-942 https://github.com/apache/incubator-spark/pull/180
