【摘要】 针对当前网络数据清理方法存在准确性和查全性差的问题,提出基于RAA的多用户网络拥塞中错误数据实时清理方法。将网络拥塞特性定义为网络延迟敏感性、网络吞吐量敏感性和网络同步突发性,将拥塞原因定义为当多个客户端共同向一个服务器请求服务时,服务器会产生大量数据流,且链路网处理性能会比数据包的传输速率要低。依据网络拥塞特性与产生因素分析,基于网络拥塞和数据包在该链路通过时延成正比理念,估计出网络拥塞链路。在拥塞链路节点上采集数据样本之后,先判断数据异常与否,假设为正常数据,则将该数据合并至中间聚集的结果集合中,反之放至错误数据集合;网络中各节点合并本身全部子节点聚集结果集合、错误数据集合,将合并结果传输至节点本身父节点;各节点将获取正常数据聚集结果与错误数据集合,并将错误数据集合中需要清理的数据清理掉。实验结果表明,该方法数据清理准确率和查全率均较高,具有一定可靠性。更多还原

【Abstract】 Currently, the method to clean network data leads to low accuracy and recall performance. Therefore, a real-time method to clean the erroneous data in multi-user network congestion based on RAA was proposed. At first, this method defined the network congestion characteristics as the network delay sensitivity, network throughput sensitivity and network synchronous burstiness. And then, our method defined the congestion reason as a fact that when some clients sent the request services to a server together, the server generated massive data streams, and the processing performance of link network might be lower than the transmission rate of data packet. According to analysis of network congestion characteristics and generation factors, network congestion links were estimated based on the idea that network congestion data was proportional to the transit delay of data packets on this link. After collecting data samples on the congested link node, we firstly judged whether the data was abnormal. If it was normal data, we amalgamated the data into the intermediate result set, otherwise we put them in the error data set. Each node in the network merged all its sub-nodes and aggregated the result set and error data set, and then the merged result was transmitted to the parent node. After that, each node obtained the normal data aggregation result and the error data set. Finally, we cleaned the data that needs to be cleaned up in the error data set. Simulation results show that the proposed method has high data clearing accuracy and recall rate. Thus, this method is reliable.更多还原