OpenFlowSDN

Implementation of a full-mesh three switch topology, using a POX controller.

The project requirements are as follows:

First of all, it is necessary to set up the Mininet virtual machine properly.

Mininet Setup

In the first place, it is necessary to run dhclient, which sets the IP of all interfaces created on the virtual machine.

sudo dhclient

We can use different PuTTy sessions to connect to the virtual machine.

Mininet Topology

The topology implemented is as the following sketch:

Source code of the topology and its description

In order to execute the topology script and create it on Mininet.

sudo python topology.py

1.Importing libraries

In this section I imported all the libraries needed to create the components of the topology above.

from mininet.net import Mininet
from mininet.node import Controller,OVSKernelSwitch
from mininet.link import TCLink
from mininet.cli import CLI
from mininet.log import setLogLevel
from mininet.node import RemoteController

2.Creating the Mininet instance

Here I created a Mininet class object and as arguments I passed the type of links and switches, respectively.

net = Mininet( link=TCLink, switch=OVSKernelSwitch )

3.Creating the hosts, switches and controller

The following lines show the creation of the 15 hosts of our topology, assigning each IP address, MAC address and their default route. Each host is named by the switch associated.

Host 1 of switch 1: h1_s1
Host 4 of switch 3: h4_s3

There are 3 LANs, one for each switch:

Switch 1 and its hosts: 10.0.0.0/8
Switch 2 and its hosts: 11.0.0.0/8
Switch 3 and its hosts: 12.0.0.0/8

There is also a pattern concerning the MAC addresses---> The 4 least significant bits of the last byte are the host number, and the 4 most significant bits of the last byte an enumeration from 0 to 2.

This is helpful to identify in a more simple way the corresponding MAC addresses.

Finally, I added the 3 switches and the controller to the net variable.

I also assigned a port(6633) and an IP address(127.0.0.1) to the remote controller.

h1_s1 = net.addHost( 'h1_s1', ip='10.0.0.1/8',mac='00:00:00:00:00:01',defaultRoute="via 10.0.0.1")
h2_s1 = net.addHost( 'h2_s1', ip='10.0.0.2/8',mac='00:00:00:00:00:02',defaultRoute="via 10.0.0.1")
h3_s1 = net.addHost( 'h3_s1', ip='10.0.0.3/8',mac='00:00:00:00:00:03',defaultRoute="via 10.0.0.1")
h4_s1 = net.addHost( 'h4_s1', ip='10.0.0.4/8',mac='00:00:00:00:00:04',defaultRoute="via 10.0.0.1")
h5_s1 = net.addHost( 'h5_s1', ip='10.0.0.5/8',mac='00:00:00:00:00:05',defaultRoute="via 10.0.0.1")

h1_s2 = net.addHost( 'h1_s2', ip='11.0.0.1/8',mac='00:00:00:00:00:11',defaultRoute="via 11.0.0.1")
h2_s2 = net.addHost( 'h2_s2', ip='11.0.0.2/8',mac='00:00:00:00:00:12',defaultRoute="via 11.0.0.1")
h3_s2 = net.addHost( 'h3_s2', ip='11.0.0.3/8',mac='00:00:00:00:00:13',defaultRoute="via 11.0.0.1")
h4_s2 = net.addHost( 'h4_s2', ip='11.0.0.4/8',mac='00:00:00:00:00:14',defaultRoute="via 11.0.0.1")
h5_s2 = net.addHost( 'h5_s2', ip='11.0.0.5/8',mac='00:00:00:00:00:15',defaultRoute="via 11.0.0.1")

h1_s3 = net.addHost( 'h1_s3', ip='12.0.0.1/8',mac='00:00:00:00:00:21',defaultRoute="via 12.0.0.1")
h2_s3 = net.addHost( 'h2_s3', ip='12.0.0.2/8',mac='00:00:00:00:00:22',defaultRoute="via 12.0.0.1")
h3_s3 = net.addHost( 'h3_s3', ip='12.0.0.3/8',mac='00:00:00:00:00:23',defaultRoute="via 12.0.0.1")
h4_s3 = net.addHost( 'h4_s3', ip='12.0.0.4/8',mac='00:00:00:00:00:24',defaultRoute="via 12.0.0.1")
h5_s3 = net.addHost( 'h5_s3', ip='12.0.0.5/8',mac='00:00:00:00:00:25',defaultRoute="via 12.0.0.1")

s1 = net.addSwitch('s1')
s2 = net.addSwitch('s2')
s3 = net.addSwitch('s3')

c1 = net.addController( 'c1', controller=RemoteController, ip='127.0.0.1', port=6633)

4.Creating links

The function of these instructions is to set the links among hosts and switches, I also passed the ports to be connected as an argument, this was necessary to clarify the right paths all over the network.

net.addLink(h1_s1, s1,port1=1,port2=1)
net.addLink(h2_s1, s1,port1=1,port2=2)
net.addLink(h3_s1, s1,port1=1,port2=3)
net.addLink(h4_s1, s1,port1=1,port2=4)
net.addLink(h5_s1, s1,port1=1,port2=5)

net.addLink(h1_s2, s2,port1=1,port2=1)
net.addLink(h2_s2, s2,port1=1,port2=2)
net.addLink(h3_s2, s2,port1=1,port2=3)
net.addLink(h4_s2, s2,port1=1,port2=4)
net.addLink(h5_s2, s2,port1=1,port2=5)

net.addLink(h1_s3, s3,port1=1,port2=1)
net.addLink(h2_s3, s3,port1=1,port2=2)
net.addLink(h3_s3, s3,port1=1,port2=3)
net.addLink(h4_s3, s3,port1=1,port2=4)
net.addLink(h5_s3, s3,port1=1,port2=5)

net.addLink(s1,s2,port1=6,port2=6)
net.addLink(s2,s3,port1=7,port2=7)	
net.addLink(s1,s3,port1=7,port2=6)

5.Starting the Mininet object and the Mininet CLI(Interface)

Once the exit command is sent in the CLI, the net.stop() command will be executed, stopping the simulation.

net.build()
net.start()
CLI( net )
net.stop()

6.Set some parameters to display information.

The setLogLevel() command will display more information about the components of the topology, showing all the components in the CLI, then the main topology function is called and executed.

if __name__ == '__main__':
	setLogLevel( 'info' )
	topology()

It is important to run the following command to clean up all the simulation environment so we can start another one without having previous components or information not deleted.

sudo mn -c

POX Controller

I coded my own POX controller to fulfill the requirements of the project. First of all, the controller must be created in the pox/ext directory.

cd pox/ext

geany controller.py

Source code of the controller and its description

In order to execute the controller, we must be in the pox directory.

cd pox

Then we can execute the controller

sudo python ./pox.py controller

1.Importing libraries

In this section I imported the libraries needed to use OpenFlow commands and instructions.

I also imported other libraries which I used to implement some other processes that will be explained.

from pox.core import core
import pox.openflow.libopenflow_01 as of
from pox.openflow.flow_table import *
from pox.lib.util import dpidToStr
import os

Program sequence

1. Start listeners

Once the controller starts, the launch() function is executed, adding 2 listeners, the first one waiting for incoming switch connections and the second one waiting for incoming packets.

def launch ():
   core.openflow.addListenerByName("ConnectionUp", _handle_ConnectionUp)
   core.openflow.addListenerByName("PacketIn", _handle_PacketIn)

2. Switches connection up

Once the topology is started, the first function executed is the one below:handleConnectionUp(event).

The function waits for the switches to get started, when they start, the controller will print "Connection up: switch Datapath ID", establishing a connection between the switches and the controller.

It also assigns to the variables s1_dpid, s2_dpid and s3_dpid the real DPID of the switches in order to work with it later.

s1_dpid = 0
s2_dpid = 0
s3_dpid = 0

def _handle_ConnectionUp(event):
   
   	global s1_dpid, s2_dpid, s3_dpid
   	print("ConnectionUp: ", dpidToStr(event.connection.dpid))
   	
   	add_queues("s1-eth7")
   	add_queues("s1-eth6")
   	add_queues("s2-eth7")
   	add_queues("s2-eth6")
   	add_queues("s3-eth7")
   	add_queues("s3-eth6")

   	for m in event.connection.features.ports:
   		if m.name == "s1-eth1":
   			s1_dpid = event.connection.dpid
   		elif m.name == "s2-eth1":
   			s2_dpid = event.connection.dpid
   		elif m.name == "s3-eth1":
   			s3_dpid = event.connection.dpid

Futhermore, I have also included in this function the creation of the output queues, so they are created and set before all the process. I pass the switch port as an argument and then since there is no option with OpenFlow 1.0 version to create queues, I used the OS module to run a ovs-vsctl command from the controller.

Concerning the output queues:

First of all, I set the link among switches as max. 100MBps links---> other-config:max-rate=100000000
The first queue, uses a bandwith of 1 MBit --->create queue other-config:min-rate=1000000 other-config:max-rate=1000000
The second queue, uses a bandwith of 2 MBit --->create queue other-config:min-rate=2000000 other-config:max-rate=2000000
The third queue, uses a bandwith of 3 MBit --->create queue other-config:min-rate=3000000 other-config:max-rate=3000000

Finally, I run this command using os.system(command) including the > /dev/null instruction to avoid displaying the command output.

def add_queues(port):
	command = 'ovs-vsctl set port '+port+' qos=@newqos -- \
--id=@newqos create QoS type=linux-htb  \
 other-config:max-rate=100000000 \
 queues:1=@1\
 queues:2=@2 \
 queues:3=@3 -- \
--id=@1 create queue other-config:min-rate=1000000 other-config:max-rate=1000000 -- \
--id=@2 create queue other-config:min-rate=2000000 other-config:max-rate=2000000 -- \
--id=@3 create queue other-config:min-rate=3000000 other-config:max-rate=3000000 > /dev/null'
	os.system(command)

3. Reception of a packet

After the switches are properly connected to the controller, the next step is to wait until a packet arrives to any of the switches.

I get the packet data by using event.parsed and I also assign the global type to switch Datapath IDs variables so I can access them from these function.

There are 3 IF conditions:

First one: if event.connection.dpid==s1_dpid: ---> If the event is on switch1
Second one: if event.connection.dpid==s2_dpid: ---> If the event is on switch2
Third one: if event.connection.dpid==s3_dpid: ---> If the event is on switch3

The content is the same in the 3 conditions.

First, we set the operation type, in this case ofp_flow_mod(), this is used to set flow rules in the flow table of the switches. Then I set the hard timeout as 30 sec. In the next line, i specify the protocol code, in this case 0x0806 ARP and then the flow will be set in the switch using the instruction (of.ofp_action_output(port = of.OFPP_ALL)), finally with event.connection.send(msg) the flow is sent to the switch. This is necessary to get the reachability through all the 3 networks.
Then, I call a function called add_ sameNetworkFlows(event, ip). Explanation below this code.
After setting all the flows inside the LANs, it is time to set them among different LANs. I call twice the function add_flows, for instance for the S1 condition, the first call is to set the flows between S1 network (10.0.0.0) and the S2 network (11.0.0.0), and the second one is to set the flow rules between S1 LAN and S3 LAN. I also pass a third argument responsible to set the destination MAC addresses properly. Explanation below.

def _handle_PacketIn (event):
	packet = event.parsed
	global s1_dpid, s2_dpid, s3_dpid
	
	if event.connection.dpid==s1_dpid:
		msg = of.ofp_flow_mod()
		msg.hard_timeout = 30
		msg.match.dl_type = 0x0806
		msg.actions.append(of.ofp_action_output(port = of.OFPP_ALL))
		event.connection.send(msg)
		
		add_sameNetworkFlows(event,"10.0.0.")
		
		add_flows(event,"10.0.0.","11.0.0.",1)
		add_flows(event,"10.0.0.","12.0.0.",2)
		
				
					
	elif event.connection.dpid==s2_dpid:
		msg = of.ofp_flow_mod()
		msg.hard_timeout = 30
		msg.match.dl_type = 0x0806
		msg.actions.append(of.ofp_action_output(port = of.OFPP_ALL))
		event.connection.send(msg)	
		
		add_sameNetworkFlows(event,"11.0.0.")
		
		add_flows(event,"11.0.0.","10.0.0.",0)
		add_flows(event,"11.0.0.","12.0.0.",2)
		
		
		
	elif event.connection.dpid==s3_dpid:
		msg = of.ofp_flow_mod()
		msg.hard_timeout = 30
		msg.match.dl_type = 0x0806
		msg.actions.append(of.ofp_action_output(port = of.OFPP_ALL))
		event.connection.send(msg)
		
		add_sameNetworkFlows(event,"12.0.0.")
		
		add_flows(event,"12.0.0.","10.0.0.",0)
		add_flows(event,"12.0.0.","11.0.0.",1)

add_sameNetworkFlows(ev,ipdst):

This function has 2 arguments, the first one is the event and the second one,the destination IP.

It is responsible to set the flows in the respective switch in order to reach the hosts of the same switch or network. Since the switch port number matches the last byte of the IP destination, I used a for loop and also specified the hard timeout and the IP code:0x0800.

Example:

add_sameNetworkFlows(event,"10.0.0.")--> This is the root address of the S1 LAN, called from the S1 condition.
from 1 to 5 , I set the myipdst variable as a string concatenation between the root address(10.0.0.) and the "i" number.
Don´t forget to assign it to the msg IP destination msg.match.nw_dst so it would fill the 5 IPs, 10.0.0.1 to the port 1, 10.0.0.2 to the port 2 ..... 10.0.0.5 to the port 5.
Finnally, the rule is installed in the switch with ev.connection.send(msg)

def add_sameNetworkFlows(ev,ipdst):
	for i in range(1,6):
			msg = of.ofp_flow_mod()
			msg.hard_timeout = 30
			msg.match.dl_type = 0x0800
			myipdst=ipdst+str(i)
			msg.match.nw_dst = myipdst
			msg.actions.append(of.ofp_action_output(port = i))
			ev.connection.send(msg)

add_Flows(ev,ipsrc,ipdst,n):

The functionailty of this block is to decide which port has to be set to reach another switch LAN. For instance if we want to add the flow rules to communicate from the LAN 10.0.0.0 to the LAN 11.0.0.0, the chosen port will be the s1-eth6, as this is the one connecting S1 and S2 LANs, and so on with all the LANs.

def add_flows(ev,ipsrc,ipdst,n):
inport=0
	if   ipsrc=="10.0.0." and ipdst=="11.0.0.":
		inport=6
	elif ipsrc=="10.0.0." and ipdst=="12.0.0.":
		inport=7
	elif ipsrc=="11.0.0." and ipdst=="10.0.0.":
		inport=6
	elif ipsrc=="11.0.0." and ipdst=="12.0.0.":
		inport=7
	elif ipsrc=="12.0.0." and ipdst=="10.0.0.":
		inport=6
	elif ipsrc=="12.0.0." and ipdst=="11.0.0.":
		inport=7

Here I am using a for loop very similar to the one in the add_sameNetworkFlows(ev,ipdst) function.

First difference: I am using the loop index and the "n" parameter to assign the IPs and the MAC addresses. If you remember the topology, the 4 most significant bits of the last byte of S1 hosts are 0, for s2 1 and for s3 2, e.g:
- h1_s1--->"00:00:00:00:00:01"
- h1_s2--->"00:00:00:00:00:11"
- h5_s3--->"00:00:00:00:00:25"
Then I assign the last IP byte(host number) exactly like in the function add_sameNetworkFlows(ev,ipdst), h1_s1 ---> j=1, h4_s2 --->j=4 ...
msg.actions.append(of.ofp_action_dl_addr.set_dst(EthAddr("00:00:00:00:00:"+str(n)+str(j)))) ---> I modify the destination MAC address in order to be able to communicate among 3 LANs,
Example:ipdst:11.0.0.3. Every time that a packet arrives having 11.0.0.3 as a destination IP, the MAC address destination will be modified by the one which is inside the function set_dst, in this case it would be "EthAddr("00:00:00:00:00:13)". This was the key for allowing me to implement correctly the 3 LANs requirement.
Second difference: Here I do not use of.ofp_action_output(port), instead I use ofp_action_enqueue(port,queue_id) in order to assign the rule to the 3 different output queues which I created previously.
- If the destination host is the host1 or host2 ---> All the packets destined to h1 or h2 will go through the port chosen previously and through the queue id 1.(1Mbit)
- If the destination host is the host3 or host4 ---> All the packets destined to h3 or h4 will go through the port chosen previously and through the queue id 2.(2Mbit)
- If the destination host is the host5 ---> All the packets destined to h5 will go through the port chosen previously and through the queue id 3.(3Mbit)

	for j in range(1,6):
		msg = of.ofp_flow_mod()
		msg.match.dl_type = 0x0800
		myipdst=ipdst + str(j)
		msg.match.nw_dst = myipdst
		msg.actions.append(of.ofp_action_dl_addr.set_dst(EthAddr("00:00:00:00:00:"+str(n)+str(j))))
		
		if str(myipdst).rsplit('.',1)[1]=="1" or str(myipdst).rsplit('.',1)[1]=="2":
			msg.actions.append(of.ofp_action_enqueue(port = inport,queue_id=1))
		elif str(myipdst).rsplit('.',1)[1]=="3" or str(myipdst).rsplit('.',1)[1]=="4":
			msg.actions.append(of.ofp_action_enqueue(port = inport,queue_id=2))
		elif str(myipdst).rsplit('.',1)[1]=="5":
			msg.actions.append(of.ofp_action_enqueue(port = inport,queue_id=3))
			
		ev.connection.send(msg)

Finally, the flow is installed by sending the message to the respective switch.

Testing the project

In this section, I am going to explain step by step how to test the project properly, checking that the requirements were met.

Running the controller

First you need to place the controller in the directory "pox/ext", afterwards you need to move to "pox" directory and execute the following command.

The controller now is listening for new switch connections and new packets.

Running the topology

In another terminal, use the following command:

Checking queues created

As I explained before, once the switches established a connection, the output queues are created. In another terminal, we can use the following the following command to check the output queues for instance in the port 6 of the switch 1:

tc class list dev s1-eth6

We can see that the link bandwitdh is 100MBps (third line) and following that, there are 3 output queues:

1Mbit queue
2Mbit queue
3Mbit queue

The first one, although it is necessarily created, it is not used in the project.

Pinging all

In order to measure the reachability of the controller, I am going to run on the mininet CLI the command pingall.

As you can see, every host can connect with each other all over the network.

Checking flow table rules

After pinging, we can use the following command in another terminal to check the rules created by the controller.

As I explained before, there are 5 rules for the hosts of the same switch or LAN, and 10 rules for the other hosts from the other 2 LANs , + 1 for the ARP protocol.

Measuring traffic

One of the requirements of the project was to have a programmable bandwith for 3 output queues, it can be tested using the iperf command as the following image:

It doesn´t matter the source of the packets but their destination. As long as the IP destination host ends in 1 or 2, it will have a 1 Mbit bandwith, if it ends in 3 or 4, 2 Mbit bandwith and if it ends in 5, 3 Mbit bandwith.

Finishing the simulation and cleaning the environment

Whenever a topology ends, we always need to run sudo mn -c to clean the environment and do not have conflicts for the next simulation.

BONUS: Dynamic routing considering same LAN

As a bonus work, I also implemented a project that unlike the main project, this one just uses 1 LAN and it sets the flows dynamically.

def _handle_PacketIn (event):
	packet = event.parsed
	ipp=packet.find("ipv4")
	
	if(isinstance(packet.next,ipv4)):
		add_flow(event,packet.next.srcip,packet.next.dstip)

This is the main difference from the main project, I analize the packet incoming, if it is an IPV4 packet, I get its source and destination IP addresses and then I just call the function add_flow, assigning these addreses to the msg, installing like this, just 1 inter-switch rule in the respective switch.

def add_flow(ev,ipsrc,ipdst):
	inport=0

	if str(ipsrc).rsplit('.',1)[0]=="10.0.0" and str(ipdst).rsplit('.',1)[0]=="10.0.1":
		inport=6
	elif str(ipsrc).rsplit('.',1)[0]=="10.0.0" and str(ipdst).rsplit('.',1)[0]=="10.0.2":
		inport=7
	elif str(ipsrc).rsplit('.',1)[0]=="10.0.1" and str(ipdst).rsplit('.',1)[0]=="10.0.0":
		inport=6
	elif str(ipsrc).rsplit('.',1)[0]=="10.0.1" and str(ipdst).rsplit('.',1)[0]=="10.0.2":
		inport=7
	elif str(ipsrc).rsplit('.',1)[0]=="10.0.2" and str(ipdst).rsplit('.',1)[0]=="10.0.0":
		inport=6
	elif str(ipsrc).rsplit('.',1)[0]=="10.0.2" and str(ipdst).rsplit('.',1)[0]=="10.0.1":
		inport=7
		
	
	msg = of.ofp_flow_mod()
	msg.idle_timeout = 0
	msg.hard_timeout = 0
	msg.match.dl_type = 0x0800
	msg.match.nw_src = ipsrc
	msg.match.nw_dst = ipdst
	
	if str(ipdst).rsplit('.',1)[1]=="1" or str(ipdst).rsplit('.',1)[1]=="2":
		msg.actions.append(of.ofp_action_enqueue(port = inport,queue_id=1))
	elif str(ipdst).rsplit('.',1)[1]=="3" or str(ipdst).rsplit('.',1)[1]=="4":
		msg.actions.append(of.ofp_action_enqueue(port = inport,queue_id=2))
	elif str(ipdst).rsplit('.',1)[1]=="5":
		msg.actions.append(of.ofp_action_enqueue(port = inport,queue_id=3))
		
	ev.connection.send(msg)

Pinging in order to set the rule

Checking flow table

As you can see, beside the rules for the same switch hosts and the ARP rule, there is just 1 rule created specifically for the hosts that were involved in the ping instruction.

Bonus summary

In summary, this bonus project works the same way than the main one(same bandwith management, same output queues...). The only difference is that all the switches are in the same LAN and the controller sets the flow rules dynamically, not in a row.

Name		Name	Last commit message	Last commit date
Latest commit History 87 Commits
bonus		bonus
images		images
project		project
README.md		README.md

flopezmontes/OpenFlowSDN

Folders and files

Latest commit

History

Repository files navigation

OpenFlowSDN

Mininet Setup

Mininet Topology

Source code of the topology and its description

1.Importing libraries

2.Creating the Mininet instance

3.Creating the hosts, switches and controller

4.Creating links

5.Starting the Mininet object and the Mininet CLI(Interface)

6.Set some parameters to display information.

POX Controller

Source code of the controller and its description

1.Importing libraries

Program sequence

1. Start listeners

2. Switches connection up

3. Reception of a packet

add_sameNetworkFlows(ev,ipdst):

add_Flows(ev,ipsrc,ipdst,n):

Testing the project

Running the controller

Running the topology

Checking queues created

Pinging all

Checking flow table rules

Measuring traffic

Finishing the simulation and cleaning the environment

BONUS: Dynamic routing considering same LAN

Pinging in order to set the rule

Checking flow table

Bonus summary

About

Resources

Stars

Watchers

Forks

Languages