...
It is recommended that you install deploy your SQL database close to database server in the same network as the EC server where workflows with Duplicate UDR agents will be running.
If you are unable to have both instances EC and the database server reside in the same machinenetwork, you can minimize the physical distance between the server machines network overhead between both servers as much as possible.
Optimizing Read And Write Operations
...
From our performance tests for SAP Hana SQL storage specifically, the use of column based table compared to row based tables show significant advantages in performance and disk space footprint.
There were unexpected issues during our tests when the memory capacity of the machine where SAP Hana was installed was too small. We experienced minimal issues when installing our SAP Hana in a machine database with a 32G 64GB memory capacity.
Cache Size and Segmentation
The Indexing Field of incoming UDRs should preferably be either an increasing sequence number or a timestamp with good locality to preserve performance of workflow executions.
For large cache sizes, it may be a good idea to split them to multiple workflows in order to preserve performance.
For large cache sizes, splitting batches of incoming UDRs into multiple transactions may improve performance of workflow runs.
Minimizing the cache size may improve the performance of the duplicate checks by the Duplicate UDR agents and the speed of Search and Delete operations using the Duplicate UDR Inspector.
...