Unicast network topology inference method

Abstract

The invention discloses a unicast network topology inference method. A three-tuple packet train measurement method is improved; the tomography information of leaf nodes is used for clustering the leaf nodes, the correlation between nodes is calculated according to two parameters of delay jitter and a packet loss rate, a network topology tree is constructed from bottom to top, the network topology structure is inferred according to the two parameters of the delay jitter and the packet loss rate, the topology inference accuracy is improved, the nodes are clustered by using the information of a layer where the nodes are, and a probe packet sending amount can be effectively reduced. The inference method disclosed by the invention can be applied to topology inference on networks of any scale and in any load condition; and when the topology of the target network is complicated, the probe packet sending amount can be greatly reduced, and the flow extra injected to the network is lessened.

Description

technical field [0001] The invention relates to the technical field of computer networks, in particular to a unicast network topology inference method. Background technique [0002] The existing network topology inference method based on ternary packet trains only estimates the network topology based on a single parameter (delay or delay jitter), and its inference accuracy is affected by the network load, and cannot be accurate when the network load is heavy. Infer the network topology. Aiming at the above problems, a method of measuring topology structure based on ternary grouping train is proposed, the leaf nodes are clustered by using the tomographic information of leaf nodes, and the correlation between nodes is calculated according to the two parameters of delay jitter and packet loss rate , construct the network topology tree from the bottom up. After the experimental verification in the NS2 simulation environment, this method effectively reduces the amount of detect...

AI Technical Summary

Problems solved by technology

[0003] Based on network tomography technology (networktomography, abbreviated as NT) based on the information collected at the edge nodes to infer the connection of intermediate nodes in the network. At present, there are mainly two types of network topology inference methods based on NT technology: one is based on multiple The measurement based on unicast requires the nodes in the network (i.e., routers) to support multicast, which has the advantage that it does not need to inject too much extra traffic into the network; the other type is based on unicast measurement, which divides nodes into multiple node pairs , to measure each node pair. Compared with multicast measurement, the additional traffic injected into the network is more, but the requirements for equipment are lower. Since it is impossible to know whether the network equipment supports multicast in advance, unicast measurement is closer to on the actual network

[0004] The commonly used unicast measurement methods mainly include immediate packet pair, quaternary packet train and triple packet train measurement methods. The immediate packet pair measurement method, which mainly infers the topology based on the packet loss rate of the link, has less packet loss The correct topology cannot be deduced in cases where

The four-tuple packet train measurement method consisting of two consecutive packet pairs adds a lot of extra traffic to the network, and does not work well when the network background traffic is heavy

Images