Intelligent analysis and prognosis-based data partitioning orchestration for optimizing database performance. Partitioning is not limited to partitioning keys established solely based on the columns of the table being partitioned, rather analysis is undertaken on dependent tables and the past behavior of fundamental data elements in the database is assessed as a means for determining the most optimal partitioning scheme. Thus, relevant information and values in the table being partitioned, as well as dependent tables and the fundamental data elements is used to determine how likely each record/row in the table is to be subjected to a data manipulation operation. The likelihood of a data manipulation operation being performed on each record serves as the basis for assigning the record to one of a plurality of partitions.