Design And Implemetation Of The Switching Algorithm In Large-Scale Optical Switching Structure

With the dramatic increase in the transmission capacity of the backbone network and constantly coming business transmitted in the in the backbone network,electronic bottleneck problem of key node switch becomes the significant cause of limiting network throughput in switched network. So researching the switching devices is the key point in improving the performance of optical transmission network.The research include two aspects: one is the high performance scalable switching architecture; on the other hand is efficient scheduling algorithms. Now when the characteristics and the application range of different switching fabrics are very explicit, studying scheduling algorithms supporting multiple QoS performance with low delay has become the focus and also the key part of constructing high-performance switching devices. This paper studies the structure of the large-capacity seven-stage Clos optical network exchange with switching capacity of up to 10 Pbps, based on which new switching and routing algorithms that accelerate data processing speed, improve the throughput, reduce business congestion are designed to alleviate the problem currently occurring.The scheduling algorithm researched in the paper is divided into three parts based on the characteristics of the system architecture and the arrival business of seven-stage Clos switch. The first problem is middle class selection strategy in the plane. The full size of the plane is filled with 512x512 ports, and the middle class is filled with 64x64 ports based. Two circular graph coloring looping algorithm was adopted to handle business. Two-direction and same-direction looping algorithms that greatly accelerate the speed of data processing were proposed on the basis. By comparing the resource consumption and the highest clock frequency, to illustrate the difference among three methods. Complete the simulation logic function module in the modelsim software. The feasibility of the proposed algorithm is verified through FPGA experiments.Secondly, expansion algorithm was proposed based on the characteristics between planes. It solves the problem of the application of the power looping algorithm with the degree of the matrix being 2.Finally, according to the diversity of business characteristics, the system of structure of business among the planes, taking into account the non-blocking, avoiding backtracking and rearrangement considering it will cost too much time for the exhaustive algorithm to distribute business to each plane, and the traffic of backbone network remains little change. Optimization algorithm that handles only incremental business matrix for business arriving later, was proposed based on incremental matrix. Contrast ratio under different traffic loads and multiple incremental scale matrix the two algorithms in delay,block rate performance. Simulation results show that: when the amount of change in business is small, the improved algorithm based on incremental business matrix indeed has better performance in reducing latency.