Network traffic scheduling system and method based on DPI and machine learning

Abstract

The invention relates to a network traffic scheduling system and method based on DPI and machine learning, and the system comprises an SDN controller based on an SDN architecture, a data analysis server, a data acquisition server, and physical equipment, and a data transmission network is formed by the interconnection of a plurality of pieces of physical equipment. The method comprises the following steps that: an SDN controller issues a command and a rule to physical equipment, and monitors a network environment at the same time; the physical equipment identifies the network access data and marks corresponding marks on the network access data, and then forwards the network access data according to the command and the rule, wherein the plurality of marks are set by the SDN controller for different network services; the data acquisition server acquires data and uploads the data to the data analysis server; the data analysis server analyzes and classifies the data through a DPI or / and machine learning mode, and returns an analysis result to the SDN controller; and the SDN controller sets or adjusts the command and the rule according to the user setting, the analysis result and the current network environment.

Description

technical field [0001] The invention relates to the technical field of computer networks, in particular to a network traffic scheduling system and method based on DPI and machine learning. Background technique [0002] Different services on the network have different requirements for bandwidth, delay, and transmission performance. In order to allocate network resources reasonably according to different services, it is necessary to analyze the data flowing through the network and realize application-level monitoring. The rules monitor the business in the global network view, and perform on-demand scheduling and allocation of network resources according to application requirements. In recent years, with the development of network programmable technology and the continuous evolution of network architecture, a software-defined open network architecture with centralized control and network programming can meet the above requirements and become a necessary technology for future ne...

AI Technical Summary

Problems solved by technology

[0003] Due to the island mode of network devices in the traditional network architecture, which only has a partial network view, it is difficult to form a global form. It is not suitable for machine learning methods, and it is difficult to integrate DPI (Deep Packet Inspection) and machine learning technology. Packet detection, business classification, network resource scheduling and other functions, the application developed by the user cannot sense the network in real time and make corresponding adjustments in time, and the DPI detection technology cannot identify the application traffic whose signature code has not been recorded in the signature database, lagging behind Due to the release of new applications, it cannot identify encrypted network data streams, which has certain limitations

Images