Guaranteeing real-time QoS according to (m,k)-firm model

This talk addresses the use of the (m,k)-firm model for providing real-time QoS guarantee in adaptive real-time systems and in Internet. (m,k)-firm model is interesting mainly for two reasons. One is that there exists many real-time applications such as audio/video packet transmission which can “naturally” tolerate some packet losses according to a (m,k)-pattern. Another is that systems dimensioned according to the worst-case for hard real-time guarantee leads to oversized resource requirement. After having pointed out the new research directions for providing systems with (m,k)-firm real-time guarantee, an example called (m,k)-WFQ and its application to video packet transmission is detailed. (m,k)-WFQ extends WFQ to integrate (m,k)-firm constraint. Its delay upper bound and performance are given.