Data Mining for Performance Antipatterns in Component Based Systems Using Run-Time and Static Analysis
Trevor Parsons, John Murphy · 2004
Current complex distributed enterprise systems with performance requirements can be difficult to design for even the most experienced system designers. Shrinking development cycles exacerbate the problem as developers are often compelled to treat performance as an afterthought, a matter of production tuning to be accomplished after the system has been coded, integrated, functionally tested and deployed. Consequently enterprise systems often suffer from severe performance problems. We propose a framework for automatically detecting and assessing the impact of poor performance design (performance antipatterns) in component based systems, using a combination of run time and static analysis. The framework consists of four modules (a monitoring module, a detection module, an assessment module and a visualization module). The framework borrows techniques from the field of Knowledge Discovery in Databases. We intend to instantiate the framework for the Enterprise Java Bean platform.