A Collaborative Mass Flow Detection Method for Programmable Networks

Abstract

The invention provides a cooperative large flow detection method oriented to a programmable network, which is characterized by comprising the following steps: the programmable switch detects and records the flow data of the inflowing programmable switch in real time; Delivers the corresponding local threshold; the programmable switch compares the detected traffic data with the local threshold, and if it is determined that the traffic data exceeds the local threshold, a report is sent to the controller, otherwise, the report is not triggered; the controller identifies the traffic in the received report Data, extract the report of the programmable switch that forwarded the flow but did not trigger the report; the controller combines the received and extracted reports of the programmable switch, counts all the values ​​of the flow, and compares the statistical value with the network global threshold Determine whether the flow data is a large flow in the whole network; the controller adjusts the local thresholds sent to each programmable switch according to the reports received in the past.

Description

technical field [0001] The invention belongs to the technical field of computer networks, in particular to a cooperative large flow detection method oriented to a programmable network. Background technique [0002] With the development and change of advanced network technologies, especially after entering the 5G era, data centers are facing various challenges, both in terms of Internet business traffic and user scale. The year-on-year increase in network data volume has brought some anomalies to the network. For example, when traffic is scheduled to collide on the same link, the link bandwidth will be occupied for a long time, resulting in traffic congestion in the egress queue and high network delay. In order to maintain the normal operation of the network, the network operator needs to identify the traffic accumulation phenomenon in the network traffic through the real-time "big flow" detection method, detect the problem behavior in the network in time, and take preventive...

AI Technical Summary

Problems solved by technology

[0003] On the one hand, in an actual network, network operators need to monitor the traffic of the entire network to ensure network security. If traffic detection is only performed at a local location, some large flows in the entire network will not be detected. to ensure network security

At present, there have been a large number of studies on large flow detection algorithms deployed on individual switches, and a few researches on large flow detection methods for the entire network cannot balance detection accuracy and overall communication overhead.

[0004] On the other hand, today's data centers have tens of thousands of concurrent flows in a short period of time, and tracking and detecting each flow without error will consume a lot of resources. Large flow detection in traditional networks relies on analyzing data packet samples, which is difficult Real-time and refined management, measurement and statistics of massive data

Although programmable switches provide new ideas for aggregating traffic statistics and identifying large traffic directly on the data plane, upgrading all traditional devices to programmable devices will bring huge budgetary and operational burdens to enterprises and Internet service providers

Images