SDN网络业务识别数据集

SDN网络业务识别系统主要面向实时数据处理和资源管理的需求以及SDN网络架构下的业务识别与分类研究较少的问题现状而建设，SDN网络业务识别系统数据包括SDN网络业务识别系统网络流数据和SDN网络业务识别系统网络时空样本数据。在搭建的SDN实验网中的路由器上使用tcpdump工具监听网络中传输的视频、web和在线游戏业务数据，捕获网络流中的原始数据包，可以得到SDN网络业务识别系统网络流数据，这些数据主要记录了多个按照网络协议规定格式的一组字节组成的数据包，三种业务的数据量分别为1.07GB、673MB和167MB。另外，在搭建的SDN实验网中将采集到的三种业务原始网络流数据经过解析、截断/填空、图像生成、规范化以及业务标签的添加等操作之后，得到具有时空特性的网络样本集，即SDN网络业务识别系统网络时空样本数据，主要记录了转换成模型可以读取的图像数据，并用于模型的训练，完成业务识别，三种业务数据量分别为9.67MB、5.20MB和1.60MB。SDN网络业务识别系统数据的总数据量为1.78GB。

The SDN network service identification system is developed to meet the demands of real-time data processing and resource management, as well as address the current situation of limited research on service identification and classification under the SDN network architecture. The datasets of this system include network flow data and spatiotemporal sample data for SDN network service identification.

The network flow data is collected by using the tcpdump tool on routers in the constructed SDN experimental network to monitor video, web and online game service data, and capture original data packets in the network flow. This type of data mainly records data packets composed of a set of bytes in the format specified by network protocols. The data volumes of the three services are 1.07 GB, 673 MB and 167 MB respectively.

In addition, after undergoing processing operations including parsing, truncation/padding, image generation, normalization and service label addition on the collected original network flow data of the three services in the built SDN experimental network, a network sample set with spatiotemporal characteristics is obtained, namely the spatiotemporal sample data for SDN network service identification. This type of data mainly records image data readable by models, and is used for model training to realize service identification. The data volumes of the three services are 9.67 MB, 5.20 MB and 1.60 MB respectively.

The total data volume of the datasets of the SDN network service identification system is 1.78 GB.

搜集汇总

背景与挑战

背景概述

该数据集面向SDN网络业务识别与分类研究，包含网络流数据和时空样本数据两类，通过捕获视频、web和在线游戏业务原始数据包并处理生成图像样本，总数据量1.75GB，用于支持通信技术领域的模型训练和业务识别任务。