Simulation technique for available bandwidth estimation

TL;DR

This paper introduces a Variable Packet Size (VPS) method for estimating available bandwidth, analyzing its accuracy, limitations, and advantages over existing techniques through simulation and real-world testing.

Contribution

It presents a novel VPS-based bandwidth measurement method, derives its applicability limits, and compares its performance with existing tools like Pathload.

Findings

VPS method accurately estimates available bandwidth within certain delay measurement bounds.

The relationship between measurement error and the number of tests is quantified.

Experimental results show VPS outperforms some existing techniques in reliability.

Abstract

The paper proposes a method for measuring available bandwidth, based on testing network packets of various sizes (Variable Packet Size method, VPS). The boundaries of applicability of the model have been found, which are based on the accuracy of measurements of packet delays, also we have derived a formula of measuring the upper limit of bandwidth. The computer simulation has been performed and relationship between the measurement error of available bandwidth and the number of measurements has been found. Experimental verification with the use of RIPE Test Box measuring system has shown that the suggested method has advantages over existing measurement techniques. Pathload utility has been chosen as an alternative technique of measurement, and to ensure reliable results statistics by SNMP agent has been withdrawn directly from the router.