Performance of traffic engineering methods in response to rapid changes of network state

A highly desirable feature of control plane technology is support for traffic engineering (TE), where traffic demands are efficiently assigned to network resources. This paper compares two major TE methods, crankback and QoS routing and their relative performance during rapid changes in network-state. The methods were evaluated in a distributed testbed using signalling (RSVP-TE with and without crank-back) and routing (OSPF and OSPF-TE) protocols. Our results show that the crankback method provides better call-blocking performance, close to the theoretical limit, than the QoS routing method regardless of the update frequency of the link state protocol. As such crankback is more capable of responding to sudden large network changes, including responding to high provisioning rates that occur in normal operation or as a result of restoration activity. The results obtained here are applicable to various connection-oriented technologies such as ATM, MPLS, SDH/SONET and optical networking and can also be implemented on different signalling and routing protocols.