Most of z/OS sites run CICS to serve transactional applications.
These CICS environments are often very complex with thousands of different transaction codes running in hundreds of CICS regions hosted in multiple production systems.
A lot of measurements providing all the needed performance metrics and allowing to analyze in detail CICS transactions elapsed time are available in SMF 110 records.
Unfortunately, the number of these records is normally very big, so it is impossible to prepare and analyze reports showing all this data.
A statistical approach allowing to focus only on the CICS transactions experiencing an abnormal elapsed time is therefore needed.
In this presentation we discuss a possible approach we successfully implemented at some customers sites.
We will also describe a reporting drill down methodology that can be used to find problems very easily and follow up on improvement or degradation situations.
Mark Cohen Austrowiek
Monitoring and Observability