Pre-booking for reduced reservation overhead in long haul wavelength-division multiplexed (WDM) networks

In future wavelength-division multiplexed (WDM) networks, traffic flows are likely to be delivered via dynamic lightpath provisioning with end-to-end quality of service (QoS) support. Typically for guaranteed services, two-way acknowledged lightpath setup protocols are preferred rather than one-way unacknowledged protocols. However, in large backbones, the overhead in two-way reservation mechanisms is high due to the latencies involved, thus limiting the support for latency-sensitive applications. The authors propose a pre-booking mechanism to overcome the high reservation overhead in traditional reactive two-way reservation protocols. It excludes lightpath acknowledgment delay from the lightpath holding time by `pre-booking´ each burst/flow based on the estimated characteristics of the arriving traffic, thus empirically improving resource utilisation. The proposed mechanism is investigated with both constant bit rate and self-similar traffic. The results demonstrate that in large networks, such as NSFNET, the proposed pre-booking mechanism is able to bring significant benefits in terms of blocking probability and lightpath bandwidth efficiency. For example, with a maximum acceptable end-to-end delay of 90 milliseconds and a 1 in 10^-4bit loss tolerance, our pre-booking mechanism yields approximately double the supported traffic load relative to the familiar wavelength-routed optical burst switching architecture