Key Technology Of Service-Perceptive Traffic Management On Internet

With the fast development of Internet service and applications, it becomes a big challenge to classify and manage the Internet traffic efficiently, especially Peer-to-Peer traffic and the encryped traffic.Combined with the state 863 project―New Generation Network with High Trustability‖, after summarizing the status quo of traffic classsification and traffic management, this dissertation mainly researches Internet traffic early recognition and flexible management. Firstly, an early recognition algorithm based on deep packet inspection named ERBDPI is proposed. Secondly, an early recognition of encryped traffic classification method based on machine learning is discussed. Last but not least, the design and implementation of the service-perceptive Internet traffic management systerm based on network processor is put forward.Main work and contributions of this dissertation are outlined as follows:An early recognition algorithm based on deep packet inspection named ERBDPI is proposed, after opening out the early appearance feature of payload signature. ERBDPI only inspects the first several packet payload signature to classify the flow, which can determine the application before the end of a flow. Testing results show that compared to traditional packet sampling method, ERBDPI can get higher classification accuracy with lower packet sampling rate. This result opens a range of new possibilities for online traffic classification.To solve the problem of encryped traffic classification, an early recognition of encryped traffic classification method based on machine learning is proposed. The method using the statistic discirminator selected of the first serval packets of a flow to classify traffic, can recognise encryped traffic accurately and effectively.The design and implementation of the service-perceptive Internet traffic management systerm based on network processor is put forward. The subsequent flow storing scheme, the flexible access control and three-step scheduling are stressed. Testing results show that the system can efficiently provide flexible traffic management guarantee for various traffic category on OC-48 links.