How would changing the COMMIT size affect the amount of memory required to build the result set for the join, which is obviously the problem here?
Also, SAP HANA doesn't "fetch records into a database buffer". All records are in memory all the time (leaving the unload capability aside).
- Lars