I have a Table Comparison that is using the Sorted input compare method. Every time it runs all the O/S memory is exhausted (we watched the memory monitor go slowly to zero). When watching the Monitor Log there is a direct correlation in the TCSort_2-OrderBy6 line with the memory usage. I change the compare method to Row-by-row then there is only minimal memory usage and the Dataflow completes, albeit rather slowly.
The Dataflow Cache type is set to In-memory. However, there isn’t much going on in the Dataflow. It uses a SQL transform as the source so there is no possibility of joins happening on the job server. There is a lookup and the lookup table is small - 200 rows.
There are rows coming in from the SQL transform. I don’t think I’m hitting the bug where the Table Comparison goes into La-La-Land when there are no input rows.
Based on what I’m seeing in the Monitor Log it looks like the sort of the compare table is taking place at the job server, not within the database. When I look at the Display Optimized SQL I see an ORDER BY that matches the columns in the Primary Key list. I’m pretty sure it isn’t counting the incoming result set from the SQL transform because the count of rows from the SQL transform is far less than the count of rows I’m seeing in the Monitor Log for the TCSort_2-OrderBy6.
The job server is 14.1.2.378 running on Linux Redhat.
eganjp 
(BOB member since 2007-09-12)