Synthetic DB2 for z/OS Statistics
At the university of Jena in Germany we invented a concept called synthetic statistics and implemented a prototype for DB2 for z/OS, which is able to generate DB2 statistics on a synthetic way only based on abstract Performance Indicators (PIs).
PIs are information, that decribe the characteristics of data and can be specified by application architects / developers (e. g. within a data model). By using PIs and the concept of synthetic statistics for example it's possible to generate synthetic statistics in a very early design phase of a new database application (or new application moduls) and evaluate its performance based on execution plans without the need of possibly huge test data generation or costly data anonymisation to be able to build necessary real statistics. Addionally there is the opportunity to reverse engineer statistics to PIs, which can also be very usefull within development and review process.
Because of the positive feedback at IDUG we talked about the concept of synthetic statistics and possible integration capabilities into Optim Query Workload Tuner with Anson Kokkat (IBM), Jason Sizto (IBM) and Georg Kistenberger (IBM). To follow up this topic, we decided to open a RFE.