一、实验目的
在实验 2 的基础上进一步熟悉 Mininet 自定义拓扑脚本，以及与损耗率相关的设
定；初步了解 Mininet 安装时自带的 POX 控制器脚本编写，测试路径损耗率。
二、实验任务
h0 向 h1 发送数据包，由于在 Mininet 脚本中设置了连接损耗率，在传输过程中
会丢失一些包，本次实验的目的是展示如何通过控制器计算路径损耗速率（h0
s0-s1-h1）。这里假设控制器预先知道网络拓扑。控制器将向 s0 和 s1 发送
flow_stats_request，当控制器接收到来自 s0 的 response 时，将特定流的数据包
数保存在 input_pkts 中，当控制器接收到来自 s1 的 response 时，将接收到特定
流的数据包数保存在 output_pkts 中，差值就是丢失的数据包数量。
基于上述拓扑，编写 Mininet 脚本，设置特定的交换机间的路径损耗速率，然后
编写 POX 控制器脚本，实现对路径的损耗率的测量。
三、实验步骤
1. 实验环境
安装了 Ubuntu 18.04.5 Desktop amd64 的虚拟机
2. 实验过程
SDNLAB 实验参考资料：https://www.sdnlab.com/15100.html
（1）新建并编辑 pox 脚本 flowstat.py：
在 pox 安装目录下（Mininet 完整安装包含了 pox）执行以下命令运行 pox 脚本
$ ./pox.py flowstat
现在一起看下 flowstat.py 的关键代码：
第 7 行开始，让 h0 ping h1，监测 s0 和 s1 之间的链路。
·如果匹配到以太网类型的包头（0x0800），并且数据包的目的 IP 地址是
192.168.123.2（对照后面 Mininet 的脚本发现是 h1）， 并且连接到控制器的数
据平面设备 id 是 s0（h0 ping h1，链路 s0-s1 上数据包是从 s0 流向 s1，s0 为
源，s1 为目的地）， 执行 input_pkts = f.packet_count，把数据包数量存入
input_pkts；
·同理，如果连接到控制器的数据平面设备 id 是 s1，执行 output_pkts =
f.packet_count，把数据包数量存入 output_pkts。
·最后求 input_pkts 和 output_pkts 的差值。一般情况下差值为正，说明链路上
数据包有损耗。
#!/usr/bin/python # Copyright 2012 William Yu # wyu@ateneo.edu # # This file is part of POX. # # POX is free software: you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation, either version 3 of the License, or # (at your option) any later version. # # POX is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with POX. If not, see <http://www.gnu.org/licenses/>. # """ This is a demonstration file created to show how to obtain flow and port statistics from OpenFlow 1.0-enabled switches. The flow statistics handler contains a summary of web-only traffic. """ # standard includes from pox.core import core from pox.lib.util import dpidToStr import pox.openflow.libopenflow_01 as of from pox.lib.addresses import IPAddr, EthAddr # include as part of the betta branch from pox.openflow.of_json import * from pox.lib.recoco import Timer import time log = core.getLogger() src_dpid = 0 dst_dpid = 0 input_pkts = 0 output_pkts = 0 def getTheTime(): #fuction to create a timestamp flock = time.localtime() then = "[%s-%s-%s" %(str(flock.tm_year),str(flock.tm_mon),str(flock.tm_mday)) if int(flock.tm_hour)<10: hrs = "0%s" % (str(flock.tm_hour)) else: hrs = str(flock.tm_hour) if int(flock.tm_min)<10: mins = "0%s" % (str(flock.tm_min)) else: mins = str(flock.tm_min) if int(flock.tm_sec)<10: secs = "0%s" % (str(flock.tm_sec)) else: secs = str(flock.tm_sec) then +="]%s.%s.%s" % (hrs,mins,secs) return then # handler for timer function that sends the requests to all the # switches connected to the controller. def _timer_func (): for connection in core.openflow._connections.values(): connection.send(of.ofp_stats_request(body=of.ofp_flow_stats_request())) connection.send(of.ofp_stats_request(body=of.ofp_port_stats_request())) log.debug("Sent %i flow/port stats request(s)", len(core.openflow._connections)) # handler to display flow statistics received in JSON format # structure of event.stats is defined by ofp_flow_stats() def _handle_flowstats_received (event): #stats = flow_stats_to_list(event.stats) #log.debug("FlowStatsReceived from %s: %s", dpidToStr(event.connection.dpid), stats) global src_dpid, dst_dpid, input_pkts, output_pkts #print "src_dpid=", dpidToStr(src_dpid), "dst_dpid=", dpidToStr(dst_dpid) for f in event.stats: if f.match.dl_type==0x0800 and f.match.nw_dst==IPAddr("192.168.123.2") and f.match.nw_tos==0x64 and event.connection.dpid==src_dpid: #print "input: ", f.byte_count, f.packet_count input_pkts = f.packet_count if f.match.dl_type==0x0800 and f.match.nw_dst==IPAddr("192.168.123.2") and f.match.nw_tos==0x64 and event.connection.dpid==dst_dpid: #print "output: ", f.byte_count, f.packet_count output_pkts = f.packet_count if input_pkts !=0: print getTheTime(), "Path Loss Rate =", (input_pkts-output_pkts)*1.0/input_pkts*100, "%" # handler to display port statistics received in JSON format def _handle_portstats_received (event): #print " <<<STATS-REPLY: Return PORT stats for Switch", event.connection.dpid,"at ",getTheTime() #for f in event.stats: #if int(f.port_no)<65534: #print " PortNo:", f.port_no, " Fwd's Pkts:", f.tx_packets, " Fwd's Bytes:", f.tx_bytes, " Rc'd Pkts:", f.rx_packets, " Rc's Bytes:", f.rx_bytes #print " PortNo:", f.port_no, " TxDrop:", f.tx_dropped, " RxDrop:", f.rx_dropped, " TxErr:", f.tx_errors, " RxErr:", f.rx_errors, " CRC:", f.rx_crc_err, " Coll:", f.collisions stats = flow_stats_to_list(event.stats) log.debug("PortStatsReceived from %s: %s", dpidToStr(event.connection.dpid), stats) def _handle_ConnectionUp (event): global src_dpid, dst_dpid print "ConnectionUp: ", dpidToStr(event.connection.dpid) for m in event.connection.features.ports: if m.name == "s0-eth0": src_dpid = event.connection.dpid elif m.name == "s1-eth0": dst_dpid = event.connection.dpid msg = of.ofp_flow_mod() msg.priority =1 msg.idle_timeout = 0 msg.match.in_port =1 msg.actions.append(of.ofp_action_output(port = of.OFPP_ALL)) event.connection.send(msg) msg = of.ofp_flow_mod() msg.priority =1 msg.idle_timeout = 0 msg.match.in_port =2 msg.actions.append(of.ofp_action_output(port = of.OFPP_ALL)) event.connection.send(msg) msg = of.ofp_flow_mod() msg.priority =10 msg.idle_timeout = 0 msg.hard_timeout = 0 msg.match.dl_type = 0x0800 msg.match.nw_tos = 0x64 msg.match.in_port=1 msg.match.nw_dst = "192.168.123.2" msg.actions.append(of.ofp_action_output(port = 2)) event.connection.send(msg) msg = of.ofp_flow_mod() msg.priority =10 msg.idle_timeout = 0 msg.hard_timeout = 0 msg.match.dl_type = 0x0800 msg.match.nw_tos = 0x64 msg.match.nw_dst = "192.168.123.1" msg.actions.append(of.ofp_action_output(port = 1)) event.connection.send(msg) # main functiont to launch the module def launch (): # attach handsers to listners core.openflow.addListenerByName("FlowStatsReceived", _handle_flowstats_received) core.openflow.addListenerByName("PortStatsReceived", _handle_portstats_received) core.openflow.addListenerByName("ConnectionUp", _handle_ConnectionUp) # timer set to execute every five seconds Timer(1, _timer_func, recurring=True)
（2）编辑 Mininet 脚本 mymininet3.py
参照拓扑图，新建并编辑 Mininet 脚本 mymininet3.py，控制器因为安装在本机，
所以需修改参考资料代码中的控制器地址为 127.0.0.1:6633。
switch.cmd( 'ovs-vsctl set-controller dp0 tcp:127.0.0.1:6633' ) switch1.cmd( 'ovs-vsctl set-controller dp1 tcp:127.0.0.1:6633' )
设置 s0 和 s1 之间链路的丢包率为 0
info( "*** Creating links " ) linkopts0=dict(bw=100, delay='1ms', loss=0) linkopts1=dict(bw=100, delay='1ms', loss=0) link0=TCLink( h0, switch, **linkopts0) link1 = TCLink( switch, switch1, **linkopts1) link2 = TCLink( h1, switch1, **linkopts0)
完整代码如下：
#!/usr/bin/python from mininet.net import Mininet from mininet.node import Node from mininet.link import TCLink from mininet.log import setLogLevel, info from threading import Timer from mininet.util import quietRun from time import sleep def myNet(cname='controller', cargs='-v ptcp:'): "Create network from scratch using Open vSwitch." info( "*** Creating nodes " ) controller = Node( 'c0', inNamespace=False ) switch = Node( 's0', inNamespace=False ) switch1 = Node( 's1', inNamespace=False ) h0 = Node( 'h0' ) h1 = Node( 'h1' ) info( "*** Creating links " ) linkopts0=dict(bw=100, delay='1ms', loss=0) linkopts1=dict(bw=100, delay='1ms', loss=0) link0=TCLink( h0, switch, **linkopts0) link1 = TCLink( switch, switch1, **linkopts1) link2 = TCLink( h1, switch1, **linkopts0) #print link0.intf1, link0.intf2 link0.intf2.setMAC("0:0:0:0:0:1") link1.intf1.setMAC("0:0:0:0:0:2") link1.intf2.setMAC("0:1:0:0:0:1") link2.intf2.setMAC("0:1:0:0:0:2") info( "*** Configuring hosts " ) h0.setIP( '192.168.123.1/24' ) h1.setIP( '192.168.123.2/24' ) info( "*** Starting network using Open vSwitch " ) switch.cmd( 'ovs-vsctl del-br dp0' ) switch.cmd( 'ovs-vsctl add-br dp0' ) switch1.cmd( 'ovs-vsctl del-br dp1' ) switch1.cmd( 'ovs-vsctl add-br dp1' ) controller.cmd( cname + ' ' + cargs + '&' ) for intf in switch.intfs.values(): print intf print switch.cmd( 'ovs-vsctl add-port dp0 %s' % intf ) for intf in switch1.intfs.values(): print intf print switch1.cmd( 'ovs-vsctl add-port dp1 %s' % intf ) # Note: controller and switch are in root namespace, and we # can connect via loopback interface switch.cmd( 'ovs-vsctl set-controller dp0 tcp:127.0.0.1:6633' ) switch1.cmd( 'ovs-vsctl set-controller dp1 tcp:127.0.0.1:6633' ) info( '*** Waiting for switch to connect to controller' ) while 'is_connected' not in quietRun( 'ovs-vsctl show' ): sleep( 1 ) info( '.' ) info( ' ' ) #info( "*** Running test " ) h0.cmdPrint( 'ping -Q 0x64 -c 20 ' + h1.IP() ) sleep( 1 ) info( "*** Stopping network " ) controller.cmd( 'kill %' + cname ) switch.cmd( 'ovs-vsctl del-br dp0' ) switch.deleteIntfs() switch1.cmd( 'ovs-vsctl del-br dp1' ) switch1.deleteIntfs() info( ' ' ) if __name__ == '__main__': setLogLevel( 'info' ) info( '*** Scratch network demo (kernel datapath) ' ) Mininet.init() myNet()
再执行命令运行 Mininet 脚本 mymininet3.py
$ sudo python mymininet3.py
Ping 默认是每 1 秒钟测一次，ping 的结果会显示一个丢包率，这里的丢包率是根
据 ping 不通的次数占总次数的百分比计算得到的。上图中由于一共 ping 了 20
次，每次都能通，所以丢包率是 0。
info( "*** Creating links
" ) linkopts0=dict(bw=100, delay='1ms', loss=0) linkopts1=dict(bw=100, delay='1ms', loss=0) link0=TCLink( h0, switch, **linkopts0) link1 = TCLink( switch, switch1, **linkopts1) link2 = TCLink( h1, switch1, **linkopts0)
观察 pox 侧的实时状态更新
平均丢包率为 0，结果符合 Mininet 脚本中设置的损耗率，也有可能出现负值，
可以认为没有丢包。
如果修改代码中 s0 和 s1 之间链路的丢包率为 10。
info( "*** Creating links " ) linkopts0=dict(bw=100, delay='1ms', loss=0) linkopts1=dict(bw=100, delay='1ms', loss=10) link0=TCLink( h0, switch, **linkopts0) link1 = TCLink( switch, switch1, **linkopts1) link2 = TCLink( h1, switch1, **linkopts0)
重新运行 Mininet 脚本 mymininet3.py，20 秒时间的 ping 过程中有 icmp_seq 为
2/4/14/16/19/20 共 3 次 ping 不通，所以丢包率计算为 30%。（我的是15%）
POX 端重新测试，会发现出现丢包现象，但是实际测量出的丢包率会有浮动，链
路的性能总体受到了限制。
四、实验心得
最初我无法在pox目录下建立flowstat.py，发现pox被锁住，需要用命令sudo chmod -R 777 pox来解锁，这步之后便可以成功建立拓扑，后续按照链接中的代码并更改关键内容得以实现，设置丢包率为0的时候，丢包率0%，设置10的时候，丢包率15%。