tcptesting

所属分类:网络
开发工具:Visual C++
文件大小:14KB
下载次数:10
上传日期:2007-12-21 16:39:01
上 传 者zhukov
说明:  NSD2 QOS surpport tcl code

文件列表:
config.sh (2788, 2007-02-22)
__MACOSX (0, 2007-02-11)
__MACOSX\._config.sh (82, 2007-02-11)
lib.sh (11331, 2007-02-21)
makeplot.sh (874, 2007-02-19)
plottest.htm (1183, 2007-02-10)
__MACOSX\._plottest.htm (82, 2007-02-10)
runtest.tcl (7348, 2007-02-22)
__MACOSX\._runtest.tcl (82, 2007-02-10)
runtests.sh (1650, 2007-02-22)
template.htm (919, 2007-02-10)
__MACOSX\._template.htm (82, 2007-02-10)
template_index.htm (1045, 2007-02-19)
__MACOSX\._template_index.htm (82, 2007-02-11)
combine.sh (1412, 2007-02-23)

ns scripts implementing testing in "Experimental Evaluation of TCP Protocols for High-Speed Networks". Li, Leith & Shorten, Trans on Networking 2007. See www.hamilton.ie/net/ for details. Installation ------------ Unzip tcptesting.zip into its own directory. Edit config.sh as needed. That's it. Configuration ------------- Edit file config.sh as required. The scripts require ns, bash and gnuplot (the latter not essential if plots not needed). The range of network conditions tested can be controlled in detail from config.sh also. If using David Wei's ns linux patch (with many high-speed algorithms), then set NSLINUX=1 in config.sh. Running a test -------------- In the directory where tcptesting.zip is installed, open a shell and type "./runtests.sh start". The scripts create a top level directory named after the shell variable CCDESCRIPTION (value is set in config.sh). Under this top level directory, three directories are created: samertt, diffrtt, friendliness. These contain the test results for flows with the same RTT, with different RTTS (to measure RTT unfairness) and with a mix of new TCP and standard TCP. A lot of files are created ! Easiest way to explore the results is via the web page "index.htm" in the top directory. Follow the links to see plots. Useful Tips ----------- Its also possible to use the command "./runtests.sh restart". This will attempt to recover from an interrupted run by only running simulations for missing test cases. This can be very handy - its possible to start out with a small number of initial test cases so that things run quickly, then expand the number of test cases by editing "config.sh" and running "./runtests.sh restart". Plots and html index files are also regenerated - so its possible to reformat the indexing files and run "./runtests.sh restart" to regenerate the new format of plots etc. The script combine.sh allows results from different algorithms to be overplotted for comparison. To Do ----- *Extend test set Changelog --------- *Fixed bug in throughput measurement on high-speed paths (integer overflow) *Added support for David Wei's ns Linux patch *Output files now compressed to save space. *Fairness plots use log scale *Fixed windowOption_ bug in runtest.tcl *Throughput stats now measured from start of second flow (for consistent comparison across flows) *Minor mods to interface scripts with experimental testbed scripts *Introduced parameters to control web traffic *Changed tcpTick to be 1ms rather than 10ms. sorted scaling of srtt value from ns *Added combine.sh script to overplot results for different algorithms

近期下载者

相关文件


收藏者