代码拉取完成,页面将自动刷新
OFLOPS: OpenFlow Operations Per Second A benchmarking suite for openflow switches. If you are looking to benchmark controllers, check out ./cbench included with this package. ------------------- SUMMARY -------------------- Oflops is a modular framework for testing and benchmarking openflow switches. It uses libpcap and raw packet writing to simultaneously emulate an openflow controller and client traffic. It ships with an existing suite of tests and users are encouraged to develop their own tests. ------------------- QUICK INSTALL -------------------- debian: % sudo apt-get install libsnmp-dev libpcap-dev fedora/centos: % sudo yum install net-snmp-devel libpcap-devel TODO: make snmp utilities optional! % sh boot.sh # if no ./configure file % ./configure --with-openflow-src-dir=/path/to/openflow % make % make docs # build API documentation in ./doc % make install ------------------- QUICK RUN : local software switch -------------------- 1) Setup your software openflow switch a) Make sure that any program that broadcasts on all interfaces, is *off*, e.g., IPv6 routing. See OpenFlow README for more examples. b) Bring up some virtual interfaces: ip link add type veth # add veth0,veth1 ip link add type veth # add veth2,veth3 ip link add type veth # add veth3,veth4 for p in 0 1 2 3 4 5 ; do ifconfig veth$p up done c) Start the software switch: ofdatapath punix:/var/run/dp0.sock -i veth0,eth2 & ofprotocol tcp:localhost:6633 unix:/var/run/dp0.sock [see openflow reference implementation INSTALL file for more detail] 2) become root oflops uses libpcap and raw sockets, so you need privs 3) Run oflops with lo as the "control" interface and the other half of the vethXX links as two "data" interfaces oflops -c lo -d veth1 -d veth3 /path/to/test/module 3a) Running oflops directly from the source directory is a PITA because of libtool. The following is a sample invocation for inside the src directory. ./oflops -c lo -d veth1 -d veth3 ./example_modules/openflow_packet_in/.libs/libof_packet_in.so ------------------- QUICK RUN : hardware switch -------------------- 1) Setup your hardware switch a) physically connect at least one ethernet cable from your switch to a dedicated ethernet interface on the box where you will run oflops, e.g., eth1. In this context, dedicated means that no other applications are using that interface. Also, some tests may require more than one data channel, so it is may be necesary to connect more cables to more interfaces, e.g., eth2. b) point your switch's controller connection to the IP oflops machine on a free port, e.g., 6633. Note the interface matching the IP address, e.g., eth0. Note that this cannot be the same interface as in (1a) unless you're doing in-band control 2) become root 3) Run oflops with eth0 as the "control" interface and eth1 and eth2 as "data" interfaces oflops -c eth0 -d eth1 -d eth2 /path/to/test/module -------------------- Existing Modules ----------------------------- 1) packet_in timing: calculate the time between when data packet arrives and corresponding packet_in is received 2) port stats rtt: calculate time between port stats request and response. -------------------- Writing New Modules ----------------------------- 1) Start by reading code in example_modules/* 2) Use ./doc/* for reference (did you `make docs`?) 3) Contribute them back to oflops :-) -------------------- Contact ----------------------------- Currently oflops does not have its own mailing list. Please send questions or comments to the authors directly or openflow-discuss@openflowswitch.org
此处可能存在不合适展示的内容,页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容,可点击提交进行申诉,我们将尽快为您处理。