Research On Routing In Hybrid Electrical/Optical Datacenter Networks

With the rapid development of cloud computing,the data center network as a cloud computing infrastructure is driven by a large number of services and applications provided by cloud computing,resulting in a significant increase in data center network traffic.In order to guarantee high quality services for cloud computing,the traditional data center network needs to be optimized,so the data center network has become a hot research object.When traditionally all electrical switching data center networks are used to handle emerging applications,the problems of small capacity and high energy consumption of the electrical switching network become more and more obvious.In order to provide higher available bandwidth and lower power consumption,new hybrid electrical/optical data center network architectures have been proposed in recent years.The dynamic configuration of the optical path takes a long time,so the routing scheme needs to consider the dynamic configuration constraints of the optical path to improve the network throughput and reduce the routing cost,so as to achieve the purpose of improving the data center network performance.This thesis first abstracts the problem of dynamic single-demand routing problem in hybrid electrical/optical data center networks,and designs an integer linear programming model to solve this problem.In order to speed up the solution,this thesis presents a heuristic algorithm.The method is roughly divided into two parts,including setting the topology map according to the constraints and using the shortest path algorithm.The simulation results show that the heuristic algorithm is similar to the optimal solution in most aspects of performance.Considering that there may be multiple demands arriving at the data center network in a short period of time,there are resource competitions among these requirements.In this case,the overall constraint of these requirements can improve the routing performance.Therefore,this thesis establishes an integer linear programming model for the dynamic multi-demand routing problem in hybrid electrical/optical data center networks.At the same time,a heuristic algorithm is proposed to improve the solution efficiency.The heuristic algorithm first sorts the requirements,and then uses the shortest path algorithm to obtain the route according to the sorting results,and finally updates the network resources.The simulation results show that the heuristic algorithm performs are better than the unsorted results in all aspects.Finally,in order to quickly find the optimal solution of the multi-demand routing problem,this thesis introduces the column generation algorithm and converts a new multidemand routing problem into the main problem and the pricing problem of column generation.In order to further improve the speed of the column generation algorithm,two heuristic algorithms are proposed to replace the sub-problem model and the branch-andbound algorithm in the column generation algorithm.From the simulation results,it can be concluded that the performance of using only one heuristic algorithm is similar to the optimal solution,while the performance of using two heuristic algorithms at the same time will have a big gap with the optimal solution.