Spare Capacity Distribution Using Exact Response-Time Analysis

Real-time systems designed for use in dynamic open environments allow applications to enter and leave the system during runtime. This leads to changing runtime scenarios where the load and the spare capacity of hardware resources is influenced by the resource demand of running applications. Flexible real-time application components, with variable temporal parameters, can adapt their timing behaviour to these changing runtime scenarios and improve both, their Quality of Service (QoS) and the utilisation of system resources. In these open systems application components are often designed independently of each other, introducing the need for system management of resources at runtime. This management requires a mechanism to distribute the available system resources among the running application components, in a way that maximises resource usage and increases QoS with respect to their importance and temporal limits. This paper introduces a runtime algorithm for the distribution of spare capacity in flexible real-time systems. The efficiency of the presented algorithm is evaluated by empirical tests and performance measurements on embedded hardware.