Overhead in Available Bandwidth Estimation Tools: Evaluation and Analysis

Abstract

The current Available Bandwidth Estimation Tools (ABET's) to perform an estimation, using probes packets are inserted into the network. The utilization These packages, makes ABET's are intrusive and consumes part of which is measuring bandwidth to noise known as "Overhead Estimation Tools" (OET); it’s can produce negative effects on measurements performed by the ABET. This paper presents a complete and comparative analysis of behavior of Available Bandwidth (av_bw), of the ABET's most representative, as well as: Abing, Diettopp, Pathload, PathChirp, Traceband, IGI, PTR, Assolo and Wbest. The study with real Internet traffic, shows the percentage of test that is a factor packets affecting two main aspects of the estimation. The first, the accuracy, and increased indicating that EOT is directly proportional to the percentage of RE, reaching up to 70% in the tool evaluated with most of 30% of Cross-Traffic (CT). And second, the techniques used to send probes packets highly influences the Estimation Time (ET), where some tools that use slops spend up to 240s to converge when there is 60% CT in the network, ensuring that the estimate this technique av_bw highly congested channel, OET as much is used, resulting in inaccuracies in measurement.