| 摘要: |
| 针对IPv6网络环境中的拓扑探测和节点推断问题,提出一种结合SRv6协议的网络层析方法SNTTI。首先改进传统Back-to-Back探测报文结构和公共路径度量指标,设计了一种带有链路状态罚项的公共路径长度度量公式,通过两次不同负载下的测量结果差异衡量网络状况,减少了链路带宽和阻塞对于探测结果的影响;接着利用SRv6承载协议的路径可编程性,通过在探测报文的SRH中引入OAM扩展位,根据返回消息的时间间隔对两种基本网络拓扑结构进行区分,同时推断出目标节点在网络中的拓扑位置;改进了基于邻居节点加入的动态拓扑构建算法,并且针对网络中部分节点不支持SRv6的情况进行了讨论,提升了拓扑推断方法的适用性;最后提出了汇聚多源探测结果的拓扑融合算法,利用子树节点的分布特征进行拓扑融合,通过计算特征序列的余弦相似度定位不同拓扑中的同一个中间节点,然后补充边的分布,进而拟合得到网络中的一般拓扑结构。实验结果表明,本文提出的方法在IPv6网络环境中具有较好的探测效果,能够适应严格SRv6以及松散SRv6条件下的拓扑探测,在树状拓扑环境下较MCPM等方法准确率提高1.42～1.85倍,在复杂配置和较大规模的一般网络拓扑下具有稳定的探测性能,精度较MCPM等方法提高1.29～1.44倍,探测效率提高1.46～1.82倍,优于现有的同类网络层析探测方法。 |
| 关键词: IPv6 网络拓扑测量 网络层析 SRv6 |
| DOI：10.19363/J.cnki.cn10-1380/tn.2025.05.05 |
| 投稿时间：2023-06-03修订日期：2023-07-20 |
| 基金项目:本课题得到工信部工业互联网创新发展工程项目“工业企业网络安全综合防护平台”（No. TC200H01V） 资助。 |
|
| Research on IPv6 Topology Inference Methods with Network Tomography |
| JIN Rui,LI Haotian,WEI Songjie |
| School of Cyber Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, China |
| Abstract: |
| In view of the issue of network topology detection and node inference in the IPv6 environment, a network tomography method SNTTI combined with SRv6 protocol is proposed. Initially, the traditional Back-to-Back probe message structure and the common path metric indicators are improved. A common path length metric formula with link state penalties is designed to measure the network condition by comparing the measurement results obtained under two different loads. This approach mitigates the influence of link bandwidth and blocking on probe results. Subsequently, leveraging the path programmability offered by the SRv6 protocol, the SNTTI method introduces an OAM extension bit in the SRH of the probe message. By distinguishing between the two fundamental network topologies based on the time intervals of return messages, it infers the target node's topological position and enhances the dynamic topology construction algorithm based on neighbor node addition. Furthermore, the case of partial nodes in the network that do not support SRv6 is discussed to improve the applicability of the topology inference method. Furthermore, a topology fusion algorithm is presented to aggregate probe results from multiple sources. This algorithm utilizes the distribution characteristics of subtree nodes for topology fusion. By calculating the cosine similarity of feature sequences, identifies the same intermediate node in different topologies and complements the distribution of edges, resulting in the approximation of general topology structures in the network. Experimental results demonstrate the effectiveness of the proposed method in IPv6 network environments. It adapts well to both strict and loose SRv6 conditions for topology probing. In tree-like topology environments, it achieves an accuracy improvement of 1.42 to 1.85 times compared to methods like MCPM. In large-scale and complexly configured general network topologies, it shows stable probing performance with accuracy improvements of 1.29 to 1.44 times and probing efficiency enhancements of 1.46 to 1.82 times compared to MCPM and outperforms existing similar network tomography methods. |
| Key words: IPv6 network topology measurement network tomography SRv6 |
