Research On Available Bandwidth Measurement Method Based On Active Mode

With the development of network technology,the existing network system is more and more large and complex,which makes it difficult to understand all kinds of network behaviors,resulting in the increasing difficulty of network management,and the challenges of network security are becoming more and more prominent.For these reasons,researchers in the field of network expect use network measurement technology to better understand network behaviors and predict the network development trend.Measuring the available bandwidth of network links is an important part of network performance measurement,its preciseness helps us to better understand the network environment,it can provide valuable information in some scenes such as routing,network congestion control and network fault location.It can also supervise the Internet service providers(ISPS)to offer a guaranteed service.Active measurement is generally used to measure the available network bandwidth.The existing network available bandwidth measurement methods have many defects,such as the requirement of constant cross traffic during the measurement,slow measurement speed,excessive measurement intrusion,low accuracy in high-speed and complex network environments.Due to these defects,the existing measurement methods cannot meet the current needs,so it is urgent to propose new available bandwidth measurement methods.In response to above issues,this paper conducts a comprehensive research on existing available bandwidth measurement methods from the perspective of algorithm principle,and analyzes the problems existing in the measurement principle.After building a highly reliable and controllable network test platform,the existing tools are put on the test platform for experimental exploration,and the performance of each available bandwidth measurement tool is analyzed from the actual operation situation.As a result,it is found that the existing available bandwidth measurement tools are difficult to meet the measurement needs from the aspects of measurement accuracy,measurement speed,and measurement intrusion.Especially in the current complex and high-speed network environment,the attribute of measurement speed is becoming more and more important.Based on this,this paper proposes a new available bandwidth measurement method based on the traditional probe gap model.This method combines the machine learning method with the probe gap model,and uses the probe flow whose inter packet gap follows the exponential distribution for active detection.This method solves the problem that the traditional method needs to re-measure when packet loss occurs.The experimental results show that this method has high accuracy and the probability of measuring error not more than 5Mbps is more than 90%.The measurement speed is fast,and it can complete a measurement in one second.The intrusion of measurement is controllable,each measurement only sends 2.4Mbit probe flow to the link to be tested,while it may cause short-term congestion of the link with low available bandwidth during the measurement.