Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
None
-
None
-
None
Description
Currently when saving a DataFrame in Spark it is persisted as upserts. Having an option to do saves natively via HFiles, as the MapReduce loader does, would be a great performance improvement for large bulk loads. The current work around to reduce the load on the regionservers would be to save to csv from Spark then load via the MapReduce loader.