| 895 | | |
| 896 | | |
| 897 | | === 1d. Deflect layer 2 ping using your controller === |
| 898 | | For this exercise we have already reserved the hosts for you. There are two |
| 899 | | layer 2 ping servers running and one host that will be used as a client. You |
| 900 | | will need to login to the client using your key. |
| 901 | | {{{ |
| 902 | | ssh -i ssh/gpousrXX_key inki@pg30.emulab.net |
| 903 | | }}} |
| 904 | | |
| 905 | | After you login run : |
| 906 | | {{{ |
| 907 | | cd pingPlus |
| 908 | | }}} |
| 909 | | |
| 910 | | Run a layer two ping using each one of your two ethernet types. What do you |
| 911 | | notice? |
| 912 | | {{{ |
| 913 | | sudo ./pingPlus 00:15:17:d4:64:c2 eth1.1750 <ETH_TYPE1> |
| 914 | | sudo ./pingPlus 00:15:17:d4:64:c2 eth1.1750 <ETH_TYPE2> |
| 915 | | }}} |
| 916 | | |
| 917 | | = IV. Exercise 3: Using Software Routers to Create Your Experiment Topology = |
| 918 | | In this exercise we are going to use a click router to bandwidth limit our |
| 919 | | traffic. For this we are going to use the protogeni host you |
| 920 | | reserved in the beginning. |
| 921 | | |
| 922 | | '''Note:''' If you are using one of the hosts pc1.gpolab.bbn.com, |
| 923 | | pc11.gpolab.bbn.com or pg30.emulab.net your node has already been |
| 924 | | reserved. Please call us to tell you how to access it. |
| 925 | | |
| 926 | | * check that your sliver is ready |
| 927 | | {{{ |
| 928 | | omni.py sliverstatus -a <AM_NICKNAME> gpousr## |
| 929 | | }}} |
| 930 | | |
| 931 | | * If your sliver is ready, run readyToLogin to figure out how to login to the |
| 932 | | node. (if your host is '''pg33''' please read the notes): |
| 933 | | {{{ |
| 934 | | readyToLogin -a <AM_NICKNAME> gposlice## |
| 935 | | }}} |
| 936 | | |
| 937 | | |
| 938 | | '''NOTE 1:''' If your node is '''pg33.emulab.net''' then run |
| 939 | | {{{ |
| 940 | | xterm -e xterm -e ssh -i ssh/gpousr##_key -F ssh/config gpousr##@pg33 & & |
| 941 | | }}} |
| 942 | | |
| 943 | | * Check to see that the software has been downloaded : |
| 944 | | {{{ |
| 945 | | [alice@protogeni1 ~]$ ls |
| 946 | | config_dir pingPlus click-2.0.1 |
| 947 | | }}} |
| 948 | | |
| 949 | | * Configure the node. When you login to the node, the data interface of the |
| 950 | | host is not configured. |
| 951 | | {{{ |
| 952 | | [alice@protogeni1 ~]$ /sbin/ifconfig |
| 953 | | }}} |
| 954 | | This should list only two interface, the loopback and the control interface. |
| 955 | | Part of the tarball that was downloaded, is a script that will configure the |
| 956 | | node, and compile the Layer 2 ping software. Run |
| 957 | | {{{ |
| 958 | | ./config_dir/configure.sh |
| 959 | | }}} |
| 960 | | The output for Alice would look like |
| 961 | | {{{ |
| 962 | | [alice@protogeni1 ~]$ ./config_dir/configure.sh |
| 963 | | DEVICE=eth1.1750 |
| 964 | | USERCTL=no |
| 965 | | VLAN=yes |
| 966 | | IPADDR=10.42.131.146 |
| 967 | | NETMASK=255.255.255.0 |
| 968 | | BOOTPROTO=static |
| 969 | | ONBOOT=yes |
| 970 | | |
| 971 | | Added VLAN with VID == 1750 to IF -:eth1:- |
| 972 | | gcc -c -o packetFunctions.o packetFunctions.c |
| 973 | | gcc -c -o pingPlusListener.o pingPlusListener.c |
| 974 | | gcc packetFunctions.o pingPlusListener.o -o pingPlusListener |
| 975 | | gcc -c -o pingPlus.o pingPlus.c |
| 976 | | gcc packetFunctions.o pingPlus.o -o pingPlus |
| 977 | | }}} |
| 978 | | Verify that the data interface is up : |
| 979 | | {{{ |
| 980 | | [alice@protogeni1 ~]$ /sbin/ifconfig |
| 981 | | }}} |
| 982 | | and note the name of the data interface, the data interface is the one that |
| 983 | | has an IP address in the 10.42.131.0/24 subnet: |
| 984 | | {{{ |
| 985 | | eth1.1750 Link encap:Ethernet HWaddr 00:1F:29:32:92:4D |
| 986 | | inet addr:10.42.131.146 Bcast:10.42.131.255 Mask:255.255.255.0 |
| 987 | | inet6 addr: fe80::21f:29ff:fe32:924d/64 Scope:Link |
| 988 | | UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 |
| 989 | | RX packets:118 errors:0 dropped:0 overruns:0 frame:0 |
| 990 | | TX packets:115 errors:0 dropped:0 overruns:0 carrier:0 |
| 991 | | collisions:0 txqueuelen:0 |
| 992 | | RX bytes:5428 (5.3 KiB) TX bytes:3482 (3.4 KiB) |
| 993 | | |
| 994 | | }}} |
| 995 | | In this case the name is 'eth1.1750'. |
| 996 | | |
| 997 | | * Configure the click router |
| 998 | | {{{ |
| 999 | | cd click-2.0.1/conf |
| 1000 | | }}} |
| 1001 | | And edit the "gec12_conf.click" file to personalize it. |
| 1002 | | * '''IF_NAME''' is the name of the interface that was configured in the |
| 1003 | | previous step |
| 1004 | | * '''MAC_ADDRESS''' is the mac address of that interface. |
| 1005 | | |
| 1006 | | * Run the click router |
| 1007 | | {{{ |
| 1008 | | cd ../userlevel/ |
| 1009 | | sudo ./click ../conf/gec12_conf.click |
| 1010 | | }}} |
| 1011 | | |
| 1012 | | * Start the new controller. First of all kill the controller that is |
| 1013 | | currently |
| 1014 | | {{{ |
| 1015 | | ps -ef |grep nox_core |
| 1016 | | sudo kill -9 <PID> |
| 1017 | | }}} |
| 1018 | | Start the new controller |
| 1019 | | {{{ |
| 1020 | | ./nox_core -i ptcp:<PORT> -d mbox |
| 1021 | | }}} |
| 1022 | | |
| 1023 | | * On your two PL hosts install iperf by running : |
| 1024 | | {{{ |
| 1025 | | sudo yum install iperf |
| 1026 | | }}} |
| 1027 | | |
| 1028 | | * And then on one host run the server |
| 1029 | | {{{ |
| 1030 | | iperf -s -i 2 |
| 1031 | | }}} |
| 1032 | | Run the client on the other host |
| 1033 | | {{{ |
| 1034 | | iperf -c <host> -i 2 |
| 1035 | | }}} |
| 1036 | | |
| 1037 | | Look how your traffic is capped. |
| | 900 | * Try pinging again |
| | 901 | {{{ |
| | 902 | ping 10.42.<IP_SUBNET_OCT>.YYY |
| | 903 | }}} |
| | 904 | |
| | 905 | Notice how the first ping packets have higher delay than the subsequent packets. |
| | 906 | This is because when the first packet goes through, the switches will ask your controller about how to handle the packet. Your controller will install a flow on the switch to handle this and all subsequent packets of the flow. The large delay reflects the roundtrip time between the switch and your controller. Subsequent packets will be forwarded directly by the switch based on the flow that your controller installed and thus their switching will be faster. |
| | 907 | |
| | 908 | === 1d. Try the Layer 2 ping using your controller === |
| | 909 | Try running the Layer 2 ping that you did before but using the <ETH_TYPE> that was assigned to you, in your OpenFlow slip. |
| | 910 | {{{ |
| | 911 | sudo /usr/local/bin/pingPlus 00:0c:29:7b:c5:b3 <interface_name> <ETH_TYPE> |
| | 912 | }}} |
| | 913 | |
