A Neural-Network to Solving the Output Contention in Packet Switching Networks

Optical packet switching based on Wavelength Division Multiplexing (WDM) offers the possibility of increased granularity and more effective use bandwidth in large capacity systems, on scale of Tb/s. The natural integration of optical packet switching in photonic technology opens up the possibility of packet switching in transparent optical channels , where the packets remain from end-to-end in the optical domain, without the necessity of optoelectronic conversion. Therefore, the optical switching must be robust enough in order to provide conditions to solve the contention between optical packets in access networks, where the traffic is less aggregated. This works presents a novel approach to solving the output contention in optical packet switching networks with synchronous switching mode. A contention controller has been designed based on the Order Statics Filters (OSF) neural-network technique with a speed up factor to achieve a real-time computation of a non blocking switching high-speed high-capacity packet switch without packet loss. A neural network, OSF, which with any of binary as input, outputs 1408 the K th largest element of the array is proposed. Overall time is constant , and does not depend upon the size of the input array, being just eleven times the processing time for a single neuron. This neural network may be used as a building block for hardware implementation of order statistic filters. Keywords: Optical packets switching (OPS), Head-of-line (HOL), Mul-tiprotocol label switching (MPLS), Order static filter (OSF), Wavelength division multiplexing (WDM), generalized multi-protocol label switching (GM-PLS), First in first out (FIFO), Pattern recognition and image processing (PRIP)