The Research And Implementation Of Available Bandwidth Estimation And Trafifc Control

With more and more applications sharing network resource, applications are more likelyto interfere with each other. Research on network measurement is of great importance tooverall performance optimization and network service quality guarantee. In the paper, weconcentrate on available bandwidth estimation and traffic control on local host. Bothtechniques are related to network measurement.The available bandwidth of an end-to-end network path is the maximum throughput thatthe path can provide to a flow without reducing the cross traffic. Available bandwidth isdifficult to be measured because it varies from time to time. Some active probing tools, suchas Spruce, IGI, TOPP, PathLoad, PathChirp, have been introduced to estimate availablebandwidth. In this paper, we first have an overview of existing probing tools. Noticed thatnone of the probing tools above has made full use of the variation of probing rate, we proposea new active probing tool basing on the concept of self-introduced congestion. Abet, thenew probing tool, is carefully conducted and analyzed. For verification and efficiencyanalysis, we perform several experiments on Abet with different scenarios. Simulations onNS2show that Abet gets accurate estimation of available bandwidth, and is robust to differentscenarios.In this context, traffic control means to monitor and control bandwidth usage ofprocesses running on local host. Traffic control allows us to limit bandwidth usage of someprocesses, thus keeping important processes from interference. In the paper, we first analyzethe requirements of traffic control, and then compare the pros and cons of different technicalsolutions. We decide to develop the tool for traffic control on WFP, and the basic idea is tocreate a callout driver for packet filtering and modification. Next we introduce the design andimplementation of the tool in detail. Compared with some existing traffic control tools, thetool allows us to control download rate as well as upload rate. And it also allows us to controlmultiple processes at the same time. At last, several experiments are conducted to validate thefeatures and performance of the tool.In conclusion, we study and develop tools for available bandwidth estimation and trafficcontrol. The tools help the network to work in an efficient way.