Software dynamics: a new measure of performance for real-time software

This paper presents an approach to use concepts from continuous dynamical systems to describe behavior of real-time software. The idea is applicable to nearly all real-time software architectures. It relies on changing deadlines and taking quantitative measurements how many deadlines are missed or what is the total time of missed deadlines. A resulting graph can be approximated by a straight line or an exponential curve, from which dynamic parameters, such as sensitivity and time constant, can be inferred.