Modular Performance Analysis of Large-scale Distributed Embedded Systems: An Industrial Case Study

Abstract

Modular Performance Analysis (MPA) is a method for worst and best case analysis of real-time systems. It has been successfully applied to medium scale systems, however, for large scale systems the computation demands increase rapidly. In this paper an approximation method is introduced that simplifies the representation of arrival curves, the abstract event stream model at the basis of MPA. The approximation avoids periodic arrival curves and permits to considerably speed up the analysis of large systems, with almost negligible loss of accuracy. The MPA method and the approximation method were applied to a case study provided by EADS Innovation works. The case study introduces an avionic incabin communication system called distributed heterogeneous communication system (HCS). The results of the analysis showed that not all requirements of the case study can be met. By consequently applying the proposed approximation method, we were able to decrease the computation time for analyzing the system from52 seconds to less than one second while the calculation results remained unaected. Furthermore, the computation demands of the approximated system stayed stable for small variations of the model, while the computation time of the non approximated system could rapidly increase by factor 10 or even become uncomputable.