Publisher: IEEE, https://ieeexplore.ieee.org/document/8713683

Abstract:
Scheduling is an operation that has a significant impact on the data forwarding
performance of OBS nodes. In the case of service differentiation, the scheduling
operation is also subject to additional controls from service differentiation supporting
mechanisms in which an arriving burst can be admitted to be scheduled depending on its
priority level and the availability of wavelengths. There are several proposed scheduling
admission control models, but it is hypothesized that the traffic of arriving bursts is
stable, the admission control is then performed, based on the amount of allocated
wavelengths and the idle state of these wavelengths. However, arriving burst traffics in
OBS networks usually fluctuate and accurate traffic forecasting will help allocate
wavelengths more efficiently, increasing the scheduling efficiency in term of increasing
bandwidth utilization and reducing burst loss. This paper proposes a model of traffic
prediction based admission control, which includes a statistical method of forecasting the
arriving burst rate for flexible wavelength allocation and a mechanism of prioritizing more
resources for high-priority bursts. The simulation results have proved the efficiency of the
proposed model in terms of the burst loss and the bandwidth utilization. A Markov chain
based analysis model is also introduced to validate the proposed model.