Research On Cloud Data Center Optical Internect Architecture Based On Recursive Compound Graph

With the development of the Internet, new applications and technologies continue to emerge and the network traffic is also showing an explosive growth momentum. Data centers, as the core of data processing and transmission, are also under increasing pressure. The defections of traditional data centers, such as small bandwidth?poor scalability and high energy consumption are exposed. At the meantime, optical switching is developed rapidly for its unique high-bandwidth characteristic. How to apply and develop the optical switching technology in the data center has become a hot issue of global concern. The application of optical switching in data centers is also moving from mere optical transmission to optical-electrical hybrid networks to future all-optical data center networks. All-optical data center network is still in the stage of theoretical research, and still faces problems such as the network architecture scalability?fault tolerance and so on.To solve these problems and challenges, the main work of this paper is to explore the evolution of data center optical interconnect architecture and rules,and propose a fault-tolerant, scalable data center optical interconnect architecture based on recursive extension. Most of the newly proposed data center optical interconnect architecture only achieves a major performance, while ignoring other aspects. How to enhance the data center in a variety of key performance, is the most important data center optical interconnect development problems in the future. By comparing the fault tolerance between different architectures, this paper reveals the outstanding characteristics of the proposed network architecture,which is not only good at scalability but also high fault tolerance. Aiming at the problem of load balancing in data center, this paper proposes a multi-path load balancing wavelength routing algorithm for Fat-tree network architecture. Based on OMNeT ++ simulation platform, the relationship between traffic volume and blocking rate in different data center network topology models and the relationship between traffic volume and network average resource utilization rate are studied. This paper introduces the Fat-tree as an example. The simulation flow of building the network topology model and implementing the minimum load routing wavelength selection algorithm is introduced. According to the analysis of simulation results, the relationship between traffic volume change and data center blocking rate in different data center network topologies. The performance of the proposed multi-path load balancing algorithm is verified by comparing the load balancing algorithm with the minimum load routing algorithm in the Fat-tree network topology with the traffic growth verification algorithm's impact on the network blocking rate.