On Spectral Analysis of the Internet Delay Space and Detecting Anomalous Routing Paths

TL;DR

This paper applies robust principal component analysis to the Internet delay space to decompose latency data into expected and anomalous components, enabling detection of routing anomalies even with subtle latency deviations.

Contribution

It introduces a novel RPCA-based method to analyze delay matrices, revealing low-dimensional structures and detecting routing anomalies effectively.

Findings

Delay space decomposes into expected latency and inflation components.

The method successfully detects routing anomalies with subtle latency increases.

Low-dimensionality of delay space relates to properties of end hosts.

Abstract

Latency is one of the most critical performance metrics for a wide range of applications. Therefore, it is important to understand the underlying mechanisms that give rise to the observed latency values and diagnose the ones that are unexpectedly high. In this paper, we study the Internet delay space via robust principal component analysis (RPCA). Using RPCA, we show that the delay space, i.e. the matrix of measured round trip times between end hosts, can be decomposed into two components - the expected latency between end hosts with respect to the current state of the Internet and the inflation on the paths between the end hosts. Using this decomposition, first we study the well-known low-dimensionality phenomena of the delay space and ask what properties of the end hosts define the dimensions. Second, using the decomposition, we develop a filtering method to detect the paths which experience unexpected latencies and identify routing anomalies. We show that our filter successfully identifies an anomalous route even when its observed latency is not obviously high in magnitude.