Class-based dynamic buffer allocation for optical burst switching networks

Optical burst switching (OBS) is a promising technology to facilitate IP-over-WDM. In order to improve the quality of services (QoS) in OBS transmission networks, an adaptive method based on linear predictive filtering is proposed to dynamically allocate the limited optical buffers to different classes of traffic. Optical buffers are allocated on a per-class and per-label basis, according to the incoming traffic intensity and priority. This buffer management strategy decreases traffic loss caused by resource contention, and gives the loss-sensitive traffic precedence to access buffers over the loss-tolerant traffic. A new system model is introduced, and simulations show that our proposed adaptive-filter-based dynamic buffer allocation method provides better QoS compliance to traffic with higher priority, and achieves a lower system weighted loss than a fixed-sized buffer system does.