Highly reliable fine-grained SDN traffic monitor architecture

Abstract

The invention relates to a highly reliable fine-grained SDN traffic monitor architecture. Forwarding and monitoring of a controller are separated due to introduction of a monitor processing mechanism independent from a specific business and the controller; a monitor processor has a three-level architecture, wherein the upper level and the lower level are respectively corresponding to the controller and a switch, and the middle level processes traffic monitor data collected by a lower switch so as to generate a traffic feature set; the upper level comprises a controller command parser and a traffic feature selector; and a selected traffic feature is determined according to a selection command delivered by the controller, the traffic feature selector performs machine learning and data mining on the traffic feature set provided by the middle level according to the selected traffic feature, and generated statistics is used as decision information of the controller so as to provide accurate, predictive network-wide data for forwarding. The architecture has the advantages of high reliability, flexibility and monitor accuracy, communicates with the controller and the switch through an application program interface, has high expansibility, and is convenient to migrate and maintain.

Description

technical field [0001] The present invention relates to the technical field of computer network monitoring architecture, in particular to a high-reliability fine-grained SDN traffic monitoring architecture, which innovates and expands the traffic monitoring architecture under software-defined networks and combines it with today's deep learning technology. Background technique [0002] Modern data center networks operate on large-scale data exchange calculations, carrying massive services such as virtual cloud computing, big data applications, data center services, and multimedia transmissions. Traditional networks also rely on network administrators to handle network transients The problem is not timely enough. For example, network congestion changes dynamically with the network load. At the same time, it is difficult for traditional networks to obtain network-wide information, which makes it difficult to provide efficient and reliable guarantees for dynamic network traffic s...

AI Technical Summary

Problems solved by technology

[0005] (1) Due to the high frequency flow statistics queue will introduce monitoring overhead in the controller

[0006] (2) Traffic monitoring and forwarding use the same flow table, which creates the problem of inflexibility. At the same time, the packet characteristics used for monitoring purposes and the packet characteristics used for forwarding purposes are not always the same or overlap

[0007] (3) Due to the limited capacity of current hardware switches, the number of flow entries is very limited, which leads to reliability problems. When the number of flow tables is too large, it is easy to cause router flow table congestion

Images