Learning POX : Network virtualization
The application name is aggregator.py.
Overview
Network virtualization is a paradigm when a physical network topology details are hidden from the user behind some intermediate entity. This beast handles events from the physical switches and translates commands received from the user application.
POX has frameworks for both sides of OpenFlow, one is the controller side which is used to control switches, and another one is the switch side which takes commands from controllers. The aggregator is a switch, allowing other controllers to connect to it and send it OpenFlow commands and so forth. But underneath, it implements this by controlling other OpenFlow switches.
It aggregates all the ports and flow tables of all the underlying switches and presents them to its own controller as if they were all part of one big switch. When the controller sends a flow table entry, the aggregator translates it to the underlying switches. When the controller requests flow statistics, the aggregator collects it from all the switches and responds with a proper combined statistics reply message.
Implementation details
Underlying switches are interconnected using flow-based GRE tunnels supported by Open vSwitch NXM extension.
Example topology
Redirect action translation
During initialization stage three flow tables are created and prepopulated with service rules on each switch controlled by the aggregator. They are:
- Table 0, responsible to multiplex packets coming from local or tunnel ports to the following two tables
- Table 1, named “Remote“, is responsible to properly redirect traffic coming from a tunnel port
- Table 2, named “Openflow“, is the table where translated flow tables are installed to
The following table represents service rules that are pre-installed to each table on each switch.
Preinstalled flows
| Table | Match | Action |
| Table 0 | Port = 3 (tunnel port) | Resubmit to Table 1 |
| Table 0 | Any | Resubmit to Table 2 |
| Table 1 | Tunnel id = 0 | Redirect to controller |
| Table 1 | Tunnel id = 1 | Redirect to port 1 |
| Table 1 | Tunnel id = 2 | Redirect to port 2 |
| Table 1 | Tunnel id = 0x7e (MALL) | Redirect to All |
| Table 1 | Tunnel id = 0x7f (MFLOOD) | Flood |
| Table 2 | Any | Redirect to controller |
Ports
The aggregator hides all underlying switches and is seen as one big switch with lots of ports thus it has to translate switch port numbers. A simple formula is used:
Aggregator port number = switch port number + MAX_PORT_NUMBER * (switch dpid - 1)
Where MAX_PORT_NUMBER is a constant that defines the maximum number of ports that switch can own. In our module it is equal to 16.
| Local port | Switch dpid | Aggregator port |
| 1 | 1 | 1 |
| 2 | 1 | 2 |
| 1 | 2 | 17 |
| 2 | 2 | 18 |
Using cookies
The cookie field is used to map flow table entries between the aggregator and underlying switches.
For this purpose a cookie is generated and assigned to a flow table entry installed to the aggregator and controlled switches. Two mapping dictionaries are maintained to store correspondence between cookies.
Facilitating Nicira Extended Match(NXM) support in POX allows to use cookie as a key when modifying and deleting flow entries on switches.
Running the aggregator
- Start mininet using “sudo python agg_net.py“
- Add GRE tunnels to switches using “sudo ./aggregator.sh 2“
- Start main POX controller “./pox.py log.level –DEBUG openflow.of_01 –port=7744 forwarding.l2_pairs“
- Start the aggregator “./pox.py log.level –DEBUG edge.aggregator –ips=172.16.0.1,172.16.0.2“
Note that OVS 2.2.90 and Ubuntu 13.10 were used for experimentation.
Mininet script (agg_net.py)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 | #!/usr/bin/python from mininet.net import Mininet from mininet.node import Controller, RemoteController, Node from mininet.cli import CLI from mininet.log import setLogLevel, info from mininet.link import Link, Intf def aggNet(): NODE1_IP='172.16.0.1' NODE2_IP='172.16.0.2' CONTROLLER_IP='127.0.0.1' net = Mininet( topo=None, build=False) net.addController( 'c0', controller=RemoteController, ip=CONTROLLER_IP, port=6633) h1 = net.addHost( 'h1', ip='10.0.0.1' ) h2 = net.addHost( 'h2', ip='10.0.0.2' ) h3 = net.addHost( 'h3', ip='10.0.0.3' ) h4 = net.addHost( 'h4', ip='10.0.0.4' ) s1 = net.addSwitch( 's1' ) s2 = net.addSwitch( 's2' ) net.addLink( h1, s1 ) net.addLink( h2, s1 ) net.addLink( h3, s2 ) net.addLink( h4, s2 ) net.start() CLI( net ) net.stop() if __name__ == '__main__': setLogLevel( 'info' ) aggNet() |
Tunnel setup script (aggregator.sh)
1 2 3 4 5 6 7 8 9 10 11 12 | #!/bin/bash num=$1 echo Adding tunnels for $num switches rmmod dummy modprobe dummy numdummies=$((num+1)) for x in $(seq 1 $1); do ifconfig dummy$x 172.16.0.$x ovs-vsctl del-port s$x tun$x 2> /dev/null ovs-vsctl add-port s$x tun$x -- set Interface tun$x type=gre \ options:remote_ip=flow options:local_ip=172.16.0.$x options:key=flow done |
