Research Of Available Bandwidth Activeme Measurement Algorithm

With the unbelievable development of the network scale, the control mechanism and behavior characteristic of Internet become more and more complicated. In the area of realizing the network's behavior characteristic and performance display, helping the network's macros copy arrangement and management, overall performance optimization, guaranteeing and improving the network service quality, and promoting the information basic structure's healthy development etc., network measurement is significant. As one of the most basic and most important performance parameters, available bandwidth takes an important part in the network performance measurement.Recently, there are huge achievements in the research of available bandwidth algorithm at home or aboard. With the theory analyze and many simulation experiments, they proposed many available bandwidth algorithms. These algorithms are mainly active measurement algorithms. By sending probe packets, the receiver can get much information of the probe packets. Then we can estimate the available bandwidth with the information. Spruce algorithm estimates the available bandwidth by doing statistics and estimating each probe packet's interval in the sender and the receiver respective. IGI algorithm regards the sending rate as the available bandwidth, when the sending packets sequence's length equals the receiving packets by adjusting the sending rate continually. SloPS algorithm adjusts the sending rate continually based on the translate delay's changing trend, and then the rate when the algorithm gets convergence is regarded as the available bandwidth. PathChirp algorithm adopts an exponentially spaced chirp probing train, which makes the delay information have strong correlation, and then it can estimate available bandwidth with less probe packets rapidly. After doing research and analyze the algorithms' principle, I do coding to bring out each algorithm, simulate the circumstance with actual network traffic packets, then make simulation experiments. However, by analyzing and comparing the results of each algorithm, we found that each one has respective defaults. For example, Spruce algorithm's measurement result is not stable. IGI algorithm measurement accuracy will decrease while the network load's rate decreasing. SloPS algorithm's convergence rate is too slow and measurement time is too long. In PathChirp algorithm, probe packets are so large that the measurement accuracy will be influenced in the next link which is after the bottleneck link. Treno algorithm measurement result reflects the TCP throughout instead of the available bandwidth.Finally, we do some research of existed available bandwidth improved algorithm. We drew some experience from the algorithms found the defaults of IGI algorithm, and proposed improved IGI algorithm. IGI algorithm measurement is accurate when the network load is high burst. Because IGI algorithm's principle just considers the situation that the intervals between two probe packets become larger. In other word, the algorithm ignores the influence that the sequence change brings to the intervals.The improved algorithm adopts recursion method to estimate the sequence length. It is a new network load estimate method, which overcomes the limitation of original algorithm effectively. Furthermore, the improved algorithm also adopts the process of the hitting phenomenon, and it is gauged as the invariant mode of packet intervals during the process of the measurement. Therefore, the proposed algorithm also improves the efficiency of the measurement.