Inferring Congestion Sharing and Path Characteristics from Packet Interarrival Times
Inferring Congestion Sharing and Path Characteristics from Packet Interarrival Times
复制标题
从数据包到达间隔时间推断拥塞共享和路径特征
DOI：
--
复制
发表时间：
2001
期刊：
影响因子：
--
通讯作者：
C. Blake
中科院分区：
文献类型：
--
作者：
D. Katabi;C. Blake
本文提出了新的非侵入性测量技术来检测共享的上游拥塞和发现瓶颈路由器的链路速度。我们的技术是完全被动的，只需要到达时间的数据包和流标识符。我们的技术用于检测共享的拥塞是基于观察，从共享瓶颈的流的聚合到达跟踪具有非常不同的统计数据，从那些不共享瓶颈。特别地，对于共享瓶颈的聚合流量，到达间隔时间的熵要低得多。此外，本文确定的到达间隔分布，使发现多个上游路由器的链路带宽的模式结构。我们验证这些想法与广泛的实验在大规模的互联网测试平台和多速率控制路由器。我们发现该方法可以检测数百个流之间的任何共享瓶颈。分类错误的数量呈指数下降的跟踪数据包。此外，该方法很好地应对了沉重的交叉流量和错误呈指数下降的交叉流量在瓶颈的分数减少。与以前的建议不同，我们的技术不注入任何新的探测流量，不需要任何发送方合作，并与任何类型的流量（UDP，TCP，或多播），和各种各样的排队纪律。该方法简单且足够快，可以实时处理超过每秒10，000个数据包的速率。
This paper presents new non-intrusive measurement techniques to detect sharing of upstream congestion and discover bottleneck router link speeds. Our techniques are completely passive and require only arrival times of packets and flow identifiers. Our technique for detecting shared congestion is based upon the observation that an aggregated arrival trace from flows that share a bottleneck has very different statistics from those that do not share a bottleneck. In particular the entropy of the inter-arrival times is much lower for aggregated traffic sharing a bottleneck. Additionally this paper identifies mode structure in the inter-arrival distribution that enables discovery of the link bandwidths of multiple upstream routers. We validate these ideas with extensive experiments on a wide-scale Internet testbed and with multiple rate controlling routers. We find that the method can detect any bottleneck sharing among hundreds of flows. The classification errors decrease exponentially in the number of traced packets. Further, the method copes well with heavy cross-traffic and the errors decrease exponentially as the fraction of cross traffic at the bottleneck decreases. Unlike prior proposals, our technique does not inject any new probe traffic, does not require any sender cooperation, and works with any type of traffic (UDP, TCP, or multicast), and a wide variety of queuing disciplines. The method is simple and fast enough to be real-time for rates beyond 10,000 packets per second.