Research On Network Avaiable Bandwidth Measurement Methods

Internet based on TCP/IP network architecture and protocol stack, is the fastest growing technology in the 20th century. Internet is increasingly becoming a highly heterogeneous complex open system. Because the Internet lacks a unified and centralized management, it is difficult for us to understand and master the behavior of the network, and it's more difficult to predict its future trends. Network measurement is an important tool to monitor and comprehend the behavior of the network, and it is also an important way to optimize and reorganize the network structure and to improve the quality of the network service. Available bandwidth measurement as an important area of network measurement, has been used in router selection, access control, TCP slow start, throughput control and so on, and become a hot issue in computer networks.This paper first introduces the base of bandwidth measurement, which contains the basic concept of bandwidth measurement, the methods of network measurement, the bandwidth measurement techniques and several popular bandwidth measurement tools. Then three typical end-to-end path available bandwidth measurement algorithms (IGI/PTR, TOPP, Pathload) based on the PGM and PRM model are elaborated. Soon after we analyze the factors affecting bandwidth estimation, and explore how these different factors to affect the accuracy of network measurement. We also briefly describe the characteristics and simulation steps of network simulation tool NS2, which we used in the experiments. On the basis of analyzing the performance and problems of the existing network available bandwidth measurement algorithms, we propose a new end-to-end available bandwidth measurement algorithm called iChirp.iChirp, which is based on the concept of self-induced congestion, is a available bandwidth measurement algorithm of PRM. iChirp uses a dynamic approximate exponential flight pattern of probes and its probing stream consists of packets of different rates. It uses the predictive value of available bandwidth as the center, from which it rates spread to the ends on approximate exponential distribution. iChirp set a critical region around the predictive value of available bandwidth. In this critical region, iChirp apply more frequently sample and higher probing packets. According to the feedback, it also can adjust probing range and critical region to accelerate the convergence. In this paper, we carried out experiments using NS2 simulation tool, and compared with the classical algorithm pathChirp with CBR and exponential On/Off cross traffic. Experiments show that iChirp provides more accurate estimation of the available bandwidth rapidly and low-intrusively.