Maximum likelihood network topology identification from edge-based unicast measurements

Abstract

Network tomography is a process for inferring "internal" link-level delay and loss performance information based on end-to-end (edge) network measurements. These methods require knowledge of the network topology; therefore a first crucial step in the tomography process is topology identification. This paper considers the problem of discovering network topology solely from host-based, unicast measurements, without internal network cooperation. First, we introduce a novel delay-based measurement scheme that does not require clock synchronization, making it more practical than other previous proposals. In contrast to methods that rely on network cooperation , our methodology has the potential to identify layer two elements (provided they are logical topology branching points and induce some measurable delay). Second, we propose a maximum penalized likelihood criterion for topology identification. This is a global optimality criterion, in contrast to other recent proposals for topology identification that employ suboptimal, pair-merging strategies. We develop a novel Markov Chain Monte Carlo (MCMC) procedure for rapid determination of the most likely topologies. The performance of our new probing scheme and identification algorithm is explored through simulation and Internet experiments.

Network Topology Discovery Based on a Finite Set of Hypotheses

“…Hierarchical clustering algorithms are based on a similarity metric, which is an end-to-end measurable variable that indicates the length of the shared path between the sender and a pair of receivers and, therefore, the proximity of each pair of leaves in the sender based tree. Several types of measurements can be used as similarity metrics: end-to-end delay covariance and shared packet losses are among the most used; instead, the techniques hereafter proposed take advantage of the packet sandwich based similarity metric [7].…”

“…Equation (1) shows that (as stated also in [7]) the packet sandwich dispersion is incremented at each link k of the shared path, provided that the following relation,involving the capacity C k of link k and the capacity C k+1 of the following link, is satisfied:…”

“…In this paper, we propose a novel approach to this task, which is based on the packet sandwich probes described in [7]. Under the hypothesis that network link capacities belong to standard well known sets (Ethernet, OC), we define a finite space of possible topological hypotheses; within such a domain, the most likely hypothesis is chosen.…”

Routing Topology Identification Based on Multiple Sources End-to-End Measurements

“…Network tomography techniques [1,2], which identify the topology only based on the end-to-end measurements information become the popular choice. The fundamental idea of logical topology identification falls in the framework of metric-induced network topologies (MINT) [3][4][5][6][7][8].…”

Network Topology Inference Based on Traceroute and Tomography

“…They proposed a method of "sandwich probe" which makes use of the queue delay of packets on shared links to reconstruct the topology using Agglomerative Tree Algorithm. For the purpose of making the method applicable to general tree structure without using a threshold, they proposed the Reversible Jump Mote Carlo Markov Chain based Expectation Maximization (EM) Algorithm [4] to infer the topology. Meng-Fu Shih and Alfred O.…”