The task was to move an existing BW system from Oracle to HANA. Before the migration I drastically reduced the database size of the system by identifying and cleaning up basis and BW tables containing obsolete data, deleting or archiving unnecessary data. This helped keeping the hardware and license costs low. (At that time HANA was still licensed on the basis of memory size.) To assure that the database is not silently growing back again with time I set up a process chain regularly carrying out these clenup tasks.
A few tricks I used: One of the biggest non-cube tables was the table storing user bookmarks, where deletion was not an option – so I wrote a program to ”compress” the XML data of the bookmarks by removing semantically unnecessary parts. I have set up an automatic monitoring on the fragmentation of HANA row store, which can be a source of wasted memory. I also used the ”hot, warm, cold data” concept of HANA to avoid data that is not regularly used (e.g. budget backup versions) occupying the memory by automatically unloding this data from memory to transparently show the ”real” operative memory requirement.