Strictness of Rate-latency Service Curves

Network Calculus (NC) offers powerful methods for performance evaluation of queueing systems, especially for the worst-case analysis of communication networks. It is often used to obtain QoS guarantees in packet switched communication systems. One issue of nowadays’ research is the applicability of NC for multiplexed flows, in particular, if the FIFO property cannot be assumed when merging the individual flows. If a node serves the different flows using another schedule than FIFO, the terms ’strict’ or ’non-strict’ service curves play an important role. In this paper, we are dealing with the problems of strict and non-strict service curves in connection with aggregate scheduling. In the literature, the strictness of the service curve of the aggregated flow is reported as a fundamental precondition to get a service curve for the single individual flows at demultiplexing, if the service node process the input flows in Non-FIFO manner. The important strictness-property is assumed to be a unique feature of the service curve alone. But we will show here that this assumption is not true in general. Only the connection with the concrete input allows to classify a service as curve strict or