Scalability and Dimensioning of Network-Capacity Measurement System using Reflecting Servers

TL;DR

This paper presents a framework for scaling and dimensioning large-scale network capacity measurement systems using TWAMP, focusing on queuelength and waiting-time analysis through queuing theory and Monte Carlo methods.

Contribution

It introduces a novel framework for dimensioning large-scale network measurement systems based on explicit queuelength and waiting-time calculations.

Findings

Framework enables effective system dimensioning.

Combines M/G/1 queuing theory with Monte Carlo integration.

Addresses scalability challenges in network capacity measurement.

Abstract

In a class of methods for measurement of available path capacity and other capacity-related metrics in a network, trains of probe packets are transmitted from a sender to a receiver across a network path, and the sequences of time stamps at sending and reception are analyzed. In large-scale implementations there may potentially be interference between the probe-packet trains corresponding to several concurrent measurement sessions, due to congestion in the network and common measurement end points. This paper outlines principles for large-scale deployments of network capacity measurement methods using standardized network functionality. Further, the paper provides an in-depth study of dimensioning and scalability challenges related to the measurement end-points of such systems. The main result is a framework for dimensioning of large-scale network capacity measurement systems based on TWAMP. The framework is based on a method for explicit calculation of queuelength and waiting-time distributions, where results from M/G/1 queuing theory are combined with Monte Carlo integration.