As a result, the impact of di®erent types of catastrophic events like worm attacks,
power outages, accidental cable or link failures, etc. on connectivity in the Internet
becomes much more severe and longer lasting. Thus, maintaining the stability of
BGP is critical for preserving connectivity, thereby ensuring the delivery of data
packets. Our approach to solving this problem is to detect the occurrence of BGP
instabilities and to prevent their propagation. The detection algorithm proposed
here, performs a statistical analysis of features extracted from BGP update message
data to °ag the onset of an instability. We perform sequential change detection on
the feature time series using a Generalized Likelihood Ratio (GLR) based hypothesis
test. In order to make the detection more robust, we also exploit the temporal
correlation between changes detected across features and across peers. After the
detection, we propose techniques to analyze the update message contents to identify
the location of the root cause event. This information can then be used to design
BGP policy rules that can help to prevent the instability propagation. Our system
is designed to function online and can be deployed easily on any BGP router. We
evaluate our system using real BGP data from periods of a number of failure events
and SSFNet simulations. We show that it is e±cient in detecting instabilities with
minimum delay and very low false detection rates.
xiv |
/2 
eetop公众号
创芯大讲堂
创芯人才网
发表于 2009-8-26 23:09:24
