Automated derivation of time bounds in uniprocessor concurrent systems

The successful development of complex real-time systems depends on analysis techniques that can accurately assess the timing properties of those systems. This paper describes a technique for deriving upper and lower bounds on the time that can elapse between two given events in an execution of a concurrent software system running on a single processor under arbitrary scheduling. The technique involves generating linear inequalities expressing conditions that must be satisfied by all executions of such a system and using integer programming methods to find appropriate solutions to the inequalities. The technique does not require construction of the state space of the system and its feasibility has been demonstrated by using an extended version of the constrained expression toolset to analyze the timing properties of some concurrent systems with very large state spaces.<
>