Large-Scale Characterization and Segmentation of Internet Path Delays With Infinite HMMs
Large-Scale Characterization and Segmentation of Internet Path Delays With Infinite HMMs
复制标题
使用无限 HMM 大规模表征和分割互联网路径延迟
作者：
Maxime Mouchet;Sandrine Vaton;T. Chonavel;E. Aben;Jasper den Hertog
往返时间是计算机网络中最常用的性能指标之一。RIPE Atlas等测量平台为研究人员和网络运营商提供了前所未有的历史互联网延迟测量数据。自动处理这些测量结果将非常有用（路径性能的统计特征、变化检测、重复模式的识别等）。人类非常擅长在网络测量中发现模式，但很难将其自动化并同时处理多个时间序列。在这篇文章中，我们介绍了一种新的模型，HDP-HMM或无限隐马尔可夫模型，其在跟踪分割的性能非常接近人类的认知。我们证明，在一个标记的数据集和RIPE Atlas和CAIDA MANIC数据，该模型比经典的混合或隐马尔可夫模型更准确地表示测得的RTT时间序列。该方法在RIPE Atlas中实现，并引入了可公开访问的Web API。GitHub上提供了一个用于探索API的交互式笔记本。
Round-Trip Times are one of the most commonly collected performance metrics in computer networks. Measurement platforms such as RIPE Atlas provide researchers and network operators with an unprecedented amount of historical Internet delay measurements. It would be very useful to process these measurements automatically (statistical characterization of path performance, change detection, recognition of recurring patterns, etc.). Humans are quite good at finding patterns in network measurements, but it can be difficult to automate this and enable many time series to be processed at the same time. In this article we introduce a new model, the HDP-HMM or infinite hidden Markov model, whose performance in trace segmentation is very close to human cognition. We demonstrate, on a labeled dataset and on RIPE Atlas and CAIDA MANIC data, that this model represents measured RTT time series much more accurately than classical mixture or hidden Markov models. This method is implemented in RIPE Atlas and we introduce the publicly accessible Web API. An interactive notebook for exploring the API is available on GitHub.