Research And Implementation Of High Perfornace Data Packet Parsing Driver For SDN Controller

The rise of mobile Internet services, cloud computing, big data and other technologies needs more flexible resource requirements of traditional networking.However, the high complexity of the existing network structure is difficult to meet these requirements. SDN (Software Defined Network) has been widely recognized by the industry. The design idea of separating data forwarding and control makes the network composition more flexible and has better flexibility. SDN has been applied to a large number of enterprise networks and large data centers.In the face of massive data processing requirements, it is a challenge to maintaining good network performances. It is a good way to solve the data packet forwarding processing capability of network equipment effectively. Therefore, we did some research and analysis of the characteristics of the equipment under the SDN architecture and related technology,summarizes the technical advantages and disadvantages related to the promotion of packet processing performance, need comprehensive performance characteristics of the technology and system, put forward a kind of efficient data packet analysis of user space driver. The main work of this paper includes the following three aspects:(1) This paper studies and analyzes the characteristics of the switch and SDN controller in SDN architecture, and puts forward the requirement of the research and design of efficient packet analysis in improving the performance of the equipment network. Then for the existing technology to improve the efficiency of data transmission has been studied and analyzed, based on traditional Linux network data forwarding processing to improve the performance of the bottleneck. Then we study the unique advantages of DPDK data forwarding plane in data processing and the formation of these dominant factors, and based on the DPDK data forwarding technology designed a framework for efficient data packet analysis and work flow driven.(2) By analyzing the steps and matching process of the data packet analysis under the OpenFlow protocol, a double array trie tree model is proposed. This packet will be the original analytical model to network protocol information and feature extraction as a node of trie tree model, thus effectively reducing the packet flow table,time, enhance the data transmission performance.(3) For the multiprocessor environment data packet analysis, in order to make full use of the processing performance of CPU, we will deal with the whole process of data packets into multiple separate threads, using CPU affinity, the thread bound to separate CPU, In this case, CPU can be dedicated to the implementation of the binding thread, eliminating the need for system thread scheduling, while the implementation of multi-threaded concurrent packet parsing work, and effectively improve the overall performance of the system.Finally,we conducted a test environment to build,using effective testing tools and methods to carry out the performance test. By comparing the existing implementation scheme and the design of this paper, the correlation performance is compared, and the analysis is carried out.