In order to be able to handle an arbitrary amount of static communication segment faults in NoC-based MPSoCs, a flexible fault tolerance mechanism has to be applied. In this contribution, we present a flexible and scalable approach for fault-tolerance in NoCs, which - in contrast to existing circumvention techniques - can in principle handle any number of static faults in the routing network. It doesn't require routing restrictions (as static routing/source routing) and can basically be combined with any static or adaptive minimal or non-minimal routing algorithm. The needed additional hardware effort is low and the increase of the time for computation of routing decisions is reasonably low as well. The presented dependability technique can work hand-in-hand with a task scheduler/mapped and is applicable in critical, mixed-critical and non-critical application scenarios.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below