A Reinforcement Learning-Based Deflection Routing Scheme for Buffer-Less OBS Networks

Abstract

Optical burst switching (OBS) is a promising switching paradigm for the next generation Internet. A buffer-less OBS network can be implemented simply and cost-effectively without the need for either wavelength converters or optical buffers which are, currently, neither cost-effective nor technologically mature. However, this type of OBS networks suffers from relatively high loss probability caused by wavelength contentions at core nodes. This issue could prevent or, at least, delay the adoption of OBS networks as a solution for the next generation optical Internet. Deflection routing is one of the contention resolution approaches that have been proposed to tackle this problem. In addition to be cost-effective, it is also efficient in reducing loss probability, especially with low and moderate traffic loads. In this paper, we propose an adaptive reinforcement learning-based deflection routing scheme (RLDRS) which focuses on the route selection issue by choosing the optimal alternative output port in terms of both loss probability and delay when deflection is performed. Moreover, RLDRS limits the number of authorized deflections of each burst in order to reduce the additional traffic caused by deflection routing and to prohibit excessive deflections. Simulation results show that RLDRS reduces effectively loss probability and outperforms shortest path deflection routing (SPDR).