Research On The Flow Control Technology Based On Multi-Core Network Processor

With the rapid development of computer networks, the rapid growth in the number of network users and continuously expand and update of network application business, the network presents the development trend of service diversity. A variety of different business competes for limited network bandwidth will inevitably lead to some key business can’t get enough network resources, and then the quality of service decrease. To improve the quality of network service, network management core technology-flow control technology’s further research is imperative.In this paper, for the needs of the network service quality in a complex network environment, the flow control techniques were studied based on multi-core network flow processor. This paper firstly introduces the QoS (Quality of Service) architecture, analyzes the network traffic control mechanisms based on differentiated service architecture, and introduce the related knowledge of network processor. Select Cavium OCTEON CN68XX network processor as the research platform.Secondly, this paper focuses on the network packet classification technology, introduces the basic principles of the existing typical network packet classification algorithms, and compares them to analyze their characteristics and shortage. An improved network packet classification algorithm based on backtracking strategy named BTHS(Backtrack HyperSplit) has been proposed. The algorithm uses backtracking and feedback method to optimize the decision tree structure, which reduces the depth of the decision tree, improves time performance of the algorithm, and verifies its effectiveness through experiment.Thirdly, the queue management and scheduling techniques were studied in this paper. The author introduced the traditional queue management and scheduling algorithms, and deeply studied the characteristics of the OCTEON network processor. Based on OCTEON network processor, the traditional queue management and scheduling algorithms have been improved, and the achievement of the improved algorithm based on OCTEON network processor has been given.Finally, we design and implement flow control system based on OCTEON network processors. OCTEON network processor uses the features of hardware to accelerate the flow control system, including the adoption of multi-core parallel processing, the mixed operation of data plane and control plane, improved pipelined architecture model and queue scheduling mechanism based on the working group. In this paper, the design and implementation of control plane and data plane in the control systems have been elaborated in detail, and a network lab environment has been set up, which verifies the effectiveness of flow control system.