Independent Monitoring for Network Survivability
Abstract
This project addresses the problem of inferential topology discovery and network performance assessment. Specifically, we invent and analyze algorithms to deduce a network's internal topological structure and performance from delay and loss data based on packets exchanged among a set of monitors. The method assumes independent of network management, i.e., with no querying of routers or routing protocol data. This is a new and deep research area, especially difficult because, as we show, end-to-end measurement data is fundamentally under-determined to solve for the network topology (in straightforward manner). Among our methods, we have several negative results, and some successes, that make significant progress toward a general solution. A prototype monitoring system has been constructed and used to collect network data. The measurement technique is "sparse active monitoring" where monitors create their own traffic, but at a very low level. We address related issues of the monitor design, data collection and storage, web-based graphical user interface for analysis, and rendering of network maps. We found that our monitors were not sufficiently well synchronized to make accurate one-way delay measurements, and we discuss a method for correcting the delay data for clock drift.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2002
- Accession Number
- ADA408446