Interplay between Call Flow Dynamics and the Dissemination of QoS Routing Updates

摘要

In this paper, the authors study the interplay between flow dynamics and Quality of Service (QoS) routing through examining its impact on call blocking probability in the context of the ATM PNNI protocol. The PNNI specification consists of a routing protocol, based upon OSPF, and a signaling protocol, based upon the ITU-T's B-ISDN signaling, i.e., Q.2931. In PNNI routing, the routing information exchanged includes link state information as well as ATM QoS state information such as maximum cell transfer delay (maxCTD), cell delay variation (CDV), and available cell rate (ACR). The exchange of routing information is done by controlled flooding. In PNNI, when a flow arrives at the entry of the network, the source switch uses its local view of the network to select a path which meets the flow's QoS requirements. If it cannot find a suitable path, the Generic Call Admission Control (GCAC) of the source switch rejects the flow. If a suitable path is found, the flow set-up procedure is invoked and every switch along the path performs Actual Connection Admission Control (ACAC) to determine whether it has the requested resources. If not, the flow is rejected. Otherwise, the resources are reserved. For very large networks, PNNI also supports recursive hierarchical routing. However due to the additional complexity of aggregating topology as well as QoS metrics, they consider only non-hierarchical networks in this study. Based on a simplified version of PNNI, they examine the relationship between the frequency of QoS state updates, the QoS-routing related control traffic overhead and the call blocking probability. For instance, if the source switch uses out-of-date information to select a path, a false blocking situation (i.e., the local routing table's view of the network does not reflect the current increased resource availability), or a false probing situation (i.e., the local routing table's view of the network does not reflect the current decreased resource availability) may occur. In addition, out-of-date information could also lead the GCAC to select false routes. They investigate these issues under varying load conditions and for different QoS state update intervals. In the remainder of this extended abstract, they present the specific problem formulation, describe their approach and report their simulation results and findings.