An Architecture for Highly Reliable Fault-Tolerant Adaptive Distributed Embedded Systems

Abstract

[eng] Critical Adaptive Distributed Embedded Systems (ADESs) will play a key role in future cyber-physical systems (CPSs), since there is an increasing demand for carrying out critical operations in unpredictable environments. A critical ADES must rely on a communication subsystem that provides highly reliable and flexible hard real-time communication services, so as for the ADES's nodes to reliably and timely perform and adapt. However it is known that nodes, some of which are vital for an ADES, are the most unreliable system's elements. Thus, ADESs also need to tolerate and recover from node faults. Here we propose a complete architecture for ADESs that includes mechanisms to deal with faults affecting the communication subsystem and the nodes, and that is based on a flexible real-time Ethernet network that supports adaptivity. We quantitatively show that this architecture and its mechanisms allow attaining high reliability levels, as they are expected from critical ADESs.