Estimating End-to-End Available Bandwidth With Noises

Available bandwidth (AB) measurement technologies have been widely applied in evaluating the efficiency of the network communication. The performance of most existing AB measurement methods relies on the quality of the input data, which inevitably contains noises during the measurement. These noises can be multisourcing from end-to-end packet transmission and measuring, and in many practical situations, are nonnegligible compared with the value of measuring data. This paper focuses on estimating the end-to-end AB in terms of packet delays from sampled data that is contaminated by noises. We propose a novel algorithm, referred to as the adaptive-threshold algorithm, that can estimate and adaptively adjust a threshold value to separate the packets affected by the queueing delays from the rest of the packets to further improve the accuracy of AB measurement. A laboratory network-based simulation has been presented to evaluate the performance of our proposed algorithm. Numerical results demonstrate that the proposed algorithm can significantly improve the accuracy of the measurement and the estimated AB is within 10% difference from the true AB.