Issues on Performance Assessment of Optical Burst Switched Networks: Burst Loss Versus Packet Loss Metrics

With the increasing interest in optical burst switching (OBS) networks, the performance assessment of this kind of networks became of particular concern. Recently, some authors suggested that burst loss was not a reliable performance assessment metric for OBS networks. Refuting this claim, this paper presents simulation results obtained for a ring network, using real tributary IPv4 packets as source for the burst assembly. It is shown that burst loss, packet loss and byte loss lead to similar results over a wide range of burst assembly scenarios and network loads, using different resource reservation schemes. Therefore, burst loss is a reliable metric and can be used for evaluation of performance of optical burst switched networks, when realistic burst assembly algorithms are considered over real traffic.