Service availability driven re-configurable embedded system design

Abstract

Introduction of re-configurable hardware into embedded systems has given a new direction to fault tolerant computing. It is now feasible to satisfy the reliability and performance constraints on demanding applications while reducing the overall cost of the system. In order to evaluate the system's performance metrics, the application's specifications are mapped and scheduled to the computing and communication resources and their dynamic re-configuration capabilities are exploited. Our automated architecture design algorithm explores the design space and selects the optimal architecture which reconfigures itself into partial functional states in such a way that the most important services as perceived by the user are always available. The system availability is evaluated on the basis of Continuous Time Markov model and user's importance of service availability in partly and fully functional states. Two multi-objective genetic algorithms have been employed for architecture optimization. © 2011 Springer-Verlag.