A Value-Based Approach to Flexible Decision Making in Real-Time Dependable Systems

This paper deals with real-time dependable systems required to achieve the best usage of the resources even under adverse circumstances as transient overloads or component failures. Various possible trade-offs among the different application requirements must be considered to allow a flexible management of system resources. On the basis of a simple formulation of the concept of value in real-time systems we propose a performability metric that represents a suitable measure of system accomplishment level. A planner component is defined which is responsible for implementing a valuedriven admission policy that maximises the performability index and hence achieves the optimal utilisation of available system resources. The numerical evaluation of a system organisation taken as an example is performed for different load sets as well as a sensitivity analysis versus the most relevant model parameters.