Reseach On The Distributed Sheduling Algorithm In Large Capacity Optical Swtiching Structure

With the increasing diversification of Internet of things, cloud computing, and multimedia business, it is more difficult to meet the user’s demand of increasing the capacity of backbone transmission network. A large number of switching fabrics play a pivotal role in the backbone transmission network. With the development of optical fiber transmission technology, improving optical switching technology is imperative. Compared with electrical switching technology, optical switching technology has more switching capacity and lower power consumption, in addition to breaking through the "electronic bottleneck".Usually traffic can be controlled in backbone network and transmission requirements can be got in advance, so switching schedule algorithm based on matrix decomposition is practical and feasible. Previously, matrix decomposition algorithm have been stagnant, and most of them can have successful decomposition but with the penalty of increasing time delay. Based on the distributed structure control and distributed scheduling algorithm, a lower time delay can be guaranteed in large-capacity optical switches. Based on large-capacity optical switching structure, the work of this thesis is as follows:First of all, large capacity switch structure and its scheduling algorithm are researched. The large capacity optical switching structure is the Multi-Stage Multi-Plane which can support multiple time slots exchanging, the FTL and AWG are the core device. In distributed control system structure, each switch plane is independent and utilizes an independent scheduler. According to the idea of the distributed scheduling, the system scheduling can be divided into two steps: balanced scheduling between planes and switching scheduling within a plane.Secondly, in this system structure, the characteristics of each layer’s business in this system has been analyzed through employing different scheduling policies in each layer. Considering business characteristics within a plane, the looping algorithm based on matrix decomposition is utilized for switching scheduling. By optimizing and developing the looping algorithm, the algorithm performance can be improved and the algorithm application range is expanded. Simulation results show that under the condition of high load rate, the time delay of employing the looping algorithm can be much lower than that of using the non-matrix decomposition algorithm, So the looping algorithm is applicable for large-capacity optical switching structure.Finally, considering business characteristics between planes, the balanced scheduling employing looping algorithm is inapplicable. When the matrix degree value is equal to a power of 2, increasing the number of the ports of the business matrix model makes the looping algorithm usable for balanced scheduling. In the case of that the matrix degree value is greater than 2, the exhaustive algorithm can be used for balanced scheduling between planes. The proportionality and throughput of the exhaustive algorithm has been analyzed. Simulation results show that for the different numbers of business load and switch planes, compared with the simple polling RR scheduling algorithm, the exhaustive algorithm has better proportionality and throughput rate.