| This work is concerned with the problem of Rate Scheduling and Delay Quality of Service (QoS) constraints for communication networks with heterogenous traffic. Several new rate scheduling algorithms based on the “Earliest-Deadline-First” scenario are presented and compared through simulations. Since their mathematical analysis is difficult, due to the interdependence of the mean waiting times of packets belonging to different traffic classes, the concept of priority queues is modified to derive models that are equivalent in delay performance to the above algorithmic schemes, and a relation between the waiting times of the two classes is thereby deduced.; Furthermore, specific categories of rate proportional servers, the generalized models of General Processor Sharing (GPS) and Latency-Rate servers, are combined to provide an easier and iterative solution to the well known problem of setting the weights for GPS, given particular QoS constraints. The deterministic and time-invariant delay QoS guarantees used are then extended to the case of probabilistic and time-varying constraints so as to cover more realistic scenarios, where the rate is changing with time and an approximation of the QoS is known, rather than the exact criteria. Toward this end, network calculus and service curves theory are combined in order to define a generalized and new “statistical Latency-Rate” class of servers and to consider the case of statistical and time varying systems with dynamic QoS. Finally, this theory is applied to the specific problem of computing the losses in a time-varying shaper that are due to a given delay constraint. |
