Systems Seminar

Efficient Estimation of End-To-End Network Properties

Prof. Eric D. Kolaczyk
Department of Mathematics and Statistics
Boston University

Abstract

It is often desirable to monitor end-to-end properties, such as loss rates or packet delays, across an entire computer network. However, active end-to-end measurement in such settings does not scale well, and so complete network-wide measurement quickly becomes infeasible. More efficient measurement strategies are therefore needed. Previous work, examining this problem from a linear algebraic perspective, has shown that for exact recovery of complete end-to-end network properties, the number of paths that need to be monitored can be reduced to approximately the number of links in the network. Here we argue that in fact measurement strategies of even greater efficiency are possible.

We begin by recasting the problem as one of statistical prediction and show that end-to-end network properties may be accurately predicted in many cases using a significantly smaller set of carefully chosen paths than needed for exact recovery. We formulate a general framework for the prediction problem, propose a simple class of predictors for standard quantities of interest (e.g., averages, totals, differences), and show that linear algebraic methods of subset selection may be used to make effective choice of which paths to measure. We explore the accuracy of the resulting methods both analytically and numerically, in the context of real network topologies of varying size. The feasibility of our methods derives from the low effective rank of routing matrices as encountered in practice, which appears to be a new observation of interest in its own right.

Bio

Professor Kolaczyk's research focuses on the statistical modeling and analysis of various types of temporal, spatial, and network data, with a particular emphasis on multiscale statistical models. His work has resulted in new methods for signal and image denoising, tomographic image reconstruction, disease mapping, high-level image analysis in land cover classification, and monitoring of computer network traffic data. Professor Kolaczyk's publications have appeared in the literatures on statistical theory and methods, engineering, astronomy, geography, and computer science. His work has been supported by various grants from the Office of Naval Research and the National Science Foundation.

Time and Place: TUE., Mar. 29, at 4 pm in 2345 Engr. Hall.       *** NOTE SPECIAL DAY, TIME, & PLACE ***

SYSTEMS SEMINAR WEB PAGE: http://homepages.cae.wisc.edu/~gubner/seminar/

File "kolaczyk.shtml" last modified Tue 15 Oct 2019, 01:45 PM, CDT
Web Page Contact: John (dot) Gubner (at) wisc (dot) edu