Systematic integration of qualitative and quantitative parameter tuning methods for improving real-time system prototypes by AI techniques

Abstract

In performance improvement, there are a number of parameter tuning plans for improving a real-time system prototype. The author has developed two knowledge-based expert systems, BDES and BIES. BDES qualitatively diagnoses or identities bottlenecks and their sources, and generates qualitative improvement plan. BIES quantitatively estimates the effects of the improvement for bottleneck and their sources on the whole queueing network. BDES and BIES assume a real-time transaction oriented concurrent software system: (TCSS) as a queueing network (QN). Performance of a TCSS can be highly improved in systematic fashion with the complementary, integrated use of qualitative reasoning and quantitative reasoning. BDES and BIES are the components of TransObj which the author developed for real-time system prototyping.<>