Performance Modelling and Analysis Using POOSL for an In-Car Navigation System
To ensure quality and performance of soft real-time embedded systems, the evaluation of their properties is needed from early phases of the design.
These systems typically allow a certain rate of deadline misses. However, as they are often analysed using hard real-time techniques, which determine hard
bounds of their performance properties, they are over-dimensioned and thus expensive. Using a case study
inspired by industrial practice, we present how to
compose a suitable model for soft real-time systems
based on the formally defined modelling language
POOSL. By means of simulations for different usage scenarios, evaluation of the timing properties of
the system is provided. Furthermore, we compare our
results with two other performance modelling techniques, which are based on analytical computation,
showing that our approach leads to a more appropriate dimensioning of soft real-time systems.
|
