Segmentation-based On-Demand Burst Rescheduling Algorithm for Optical Burst Switched Networks

Abstract

Optical Burst Switching (OBS) is a type of switching technology capable of harnessing the Wavelength Division Multiplexing (WDM) capability by aggregating different client packets into larger packets known as bursts. OBS is a promising candidate for near-future all-optical networks. However, large burst loss at the core node due to contention is a key issue and it is caused by the bufferless nature of OBS architecture. This paper focuses on scheduling as a means to reduce contention and improve the overall network performance in terms of loss and throughput. Two important design criteria for scheduling algorithms are speed and link utilization. These criteria are not supported in current scheduling algorithms due to the trade-off between control packet processing speed and efficient utilization of the bandwidth resource. In this paper, a Segmentation-based On-Demand Burst Rescheduling Algorithm (SODBRA) is proposed and evaluated. In SODBRA, the aforementioned trade-off is balanced by hybridizing First-Fit Unscheduled Channel (FFUC) with void filling scheme. The proposed scheduling algorithm was evaluated using the NCTUns 6.0 event driven simulator and emulator. The results obtained show that SODBRA does reduce burst loss probability, increase network throughput and outperforms most of the existing scheduling schemes especially as the load and the number of wavelength increases.