| With the rapid development of Internet applications, cloud computing becomes more and more popular. Date Centers(DC), as the key infrastructures of cloud computing, have been a hot research topic both in industry and academia. Data Center Networks(DCN), as the key switching network within data centers,connect a large number of distributed servers. Traditional data center network has a hierarchical tree structure, using electrical packet switches. However the traditional DCNs suffer the following problems:low bisection bandwidth, poor scalability, high cost and power consumption, long packet delay and so on. The hybrid DCNs and all optical DCNs introduced Micro Electric Mechanic System(MEMS) based optical circuit switch technology into DCNs,which can increase the network capacity and reduce the power consumption, but they also have some problems:high cost and long packet delay.To solve the above problems,we proposed hybrid DCN with optical wavelength routing and electrical packet switching. It consists of Cycle-Arrayed Waveguide Grating Router(C-AWGR) based optical wavelength routing network and electrical ether packet switching network. C-AWGR is used as a building block in the optical wavelength routing network,which has the function of cyclic wavelength routing. The optical wavelength routing network can efficiently route data to the destination without any configurations. Compared with other typical DCNs, the numerical results show that the proposed DCN can reduce the cost and power consumption of DCNs, especially in a large scale network. We use the means of mathematical reasoning and establish simulation analysis model to evaluate the performance of proposed architecture and the results show that it is able to meet the bandwidth demand of burst traffic.SDN(Software Defined Network) separates the control plane and data plane. We proposed SDN based centralized traffic management and queue scheduling mechanism to be the routing strategy for the hybrid DCN architecture. Compared to traditional random early detection algorithm, SDN based centralized management of traffic queue scheduling mechanism has global state and scheduling information. In order to evaluate network performance we establish a simulation analysis model, and simulation results show that the proposed mechanism can reduce the packet drop rate and improve network throughput. |