Application performance analysis tooling

Performance analysis (run-time) tools are required to identify performance bottlenecks & other critical statistics early in the software development life cycle where they are cheaper to find & fix. They help developers tune an application for better performance by displaying details about how processing time is spent. Run-time data gathered should be integrated with source analysis to produce graphical views depicting the execution performance of the code base.

The Performance Analyzer in TK4.2 is not being carried forward into TK4.6. While it has proven buggy and unreliable, it provides the foundational requirements of a PA tool and any functionality in the existing PA should be considered requirements for 4.6 but should not be considered as necessarily a complete list of requirements. The wider user community needs to provide input. Highlighting specific functionality desired:

• Breakdown of time spent in module at an increasingly granular level (independent horizontal & vertical zooming) to identify hot spots in code – not just at the .so level, but at the source component and function level, as well as at a specific section of code (a loop for instance) within a function.
• Breakdown of resource consumption – CPU usage, I/O activity (by program, by ID), fixed/pool usage, heap/stack usage, etc (data available via data collection runs but with a viable unit test tool that is easier to run & results are easier to read)
• Includes analysis of library functions
• Merging trace files to create a conglomerate profile of multiple transactions/ECBs through the package