A sdn broadcast processing method based on arp event trigger agent

Abstract

The invention proposes an SDN broadcast processing method based on ARP event triggering agent. In this method, the controller is used as an agent to process ARP broadcast packets, and the storage mode of host state information, host information initialization, and host information real-time update mechanism are proposed to ensure that once a host is successfully connected to the network, it will be perceived, and any behavior of the host will cause its state Changes in information can update the host status information in time to ensure the integrity and real-time performance of host information. As a real-time ARP broadcast packet processing strategy, the present invention is applicable to other network models with centralized control characteristics, and the method has the advantages of reducing data plane traffic, reducing control plane traffic, reducing controller load, reducing host end load, and the like.

Description

technical field [0001] With the rapid development of Internet technology, the number of Internet applications and users has increased rapidly, and the existing TCP / IP-based Internet has gradually exposed many problems and drawbacks. Many countries have proposed next-generation Internet plans one after another, and software-defined networking has emerged as the times require. In recent years, it has attracted widespread attention from academia and the industry. Its typical example, OpenFlow, has been deployed in the large-scale experiments of GENI in the United States, JGN2plus in Japan, and OFELIA in the European Union. in the test bed. At the end of 2011, Deutsche Telekom, Google, Microsoft, Facebook, Verizon and Yahoo jointly established the Open Networking Foundation, aiming to promote the development of software-defined networks in a standardized way. More than 80 companies have participated in it so far. Recently, Google also announced that all of its global data center ...

AI Technical Summary

Problems solved by technology

[0012] (1) The controller is overloaded, and any broadcast data packet will be flooded to all switches in the network, and a packet-in event will be triggered on each switch. As the network scale expands, and the number of broadcast packets increases , will create a huge control burden on the controller;

[0013] (2) The bandwidth of the data plane is wasted. Any broadcast data packet is flooded in the entire network, which not only increases the processing burden of the switch, but more importantly, wastes a lot of precious bandwidth resources;

[0014] (3) There should be no loops in the link. Ethernet is very sensitive to loops and can easily cause broadcast storms. Most of the existing solutions use Spanning Tree Protocol to logically transform the network into a tree topology to avoid loops. However, the spanning tree protocol causes a waste of network bandwidth, limits network throughput, and reduces network robustness. Once a network link fails, there will be no backup. More importantly, when a switch fails, the process of spanning tree recalculation is usually It lasts for 30-50 seconds, which may affect normal network communication and even cause network paralysis;

[0015] (4) Increase the host-side load. The host network card receives all the broadcast packets in the network, and the CPU judges whether it is a data packet sent to the machine, and then discards or responds, wasting the computing and storage resources of the machine.

Images