Frame-based bandwidth allocation for Agile All-Photonic Networks

Abstract

The term "agility" in optical networks describes the ability to deploy bandwidth on demand at fine granularity that allows carriers to provision and deploy services rapidly. An overlaid star all-photonic WDM network, called the Agile All-Photonic Network (AAPN), can provide such agility through multiplexing over each wavelength in the time domain. The AAPN consists of a number of hybrid photonic/electronic edge nodes connected together via a number of optical core nodes. Each of the core nodes consists of a number of buffer-less transparent photonic space switches, one for each wavelength, and links to all edge nodes via optical fibers. Each edge node aggregates traffic in buffers, segments it as slots (i.e. units), and transmits the slots in photonic form across the network. For bandwidth sharing, the core nodes are required to perform reconfiguration periodically according to the traffic demand information sent by the edge nodes. In this thesis, it is studied how the core nodes distribute the bandwidth to adapt to the traffic in the edge nodes and how the edge nodes perform routing in the AAPN so that the traffic is less congested and can reach the destination with short delay. For the former problem, the alternating projections method and the quick Birkhoff-von Neumann decomposition method are proposed which provide a simple and efficient bandwidth allocation scheme for the AAPN core nodes. For the second problem, a routing method is proposed which provides a trade-off between the delay and the blocking probability, i.e. the method improves the delay performance of the network dramatically under the low load while not affecting the blocking performance at high load.