【摘要】 访问超点的实时获取有利于管理者更好地掌控网络。在高速网络条件下,访问超点检测的难点在于大流量给计算带来的压力。基于40 Gbit/s带宽的网络环境,从精度分析角度对比了流记录统计算法和基于估值原理的估算类超点检测算法。流记录统计算法采用基于重尾分布的阈值模型加以改进,估算类超点检测算法采用SRLA在GPU环境下实现,研究结果表明,估算类超点检测算法明显具有更高的检测精度和更大的改进空间,比流记录统计算法更适合被用于高速网络超点实时检测。更多还原

【Abstract】 Real-time access hyper-point helps managers better control the network. Under high-speed network conditions, the difficulty in accessing hyper-point detection lies in the pressure exerted by large traffic on computing. Based on the 40 Gbit/s bandwidth network environment, from the perspective of accuracy analysis, the flow record statistics algorithm and the estimation-based superpoint detection algorithm were compared. The flow record statistical algorithm was improved by the threshold model based on heavy-tailed distribution. The estimation-based superpoint detection algorithm was implemented in the GPU environment by SRLA. The research results show that the estimation-based hyper-point detection algorithm which has obviously higher accuracy and more room for improvement is more suitable for real-time detection of hyper-point in high-speed network than the flow record statistical algorithm.更多还原