The aim of this project is to offer a centralized solution in a Software Defined Network environment for solving the First Hop Redundancy (FHR) issue. It has been realized using the Floodlight Java framework which implements OpenFlow controllers.
Floodlight provides itself an implementation of some basic operations performed by controllers. We added our custom module in the pipeline adding the line it.unipi.ce.anaws.vrrm.VirtualRouterRedundancyManager in two configuration files just at the end of the list:
src/main/resources/META-INF/services/net.floodlight.core.module.IFloodlightModule;src/main/resources/floodlightdefault.properties.
Running the application we will make the controller available for connections on 127.0.0.1:6653.
The network is simulated through the mininet tool which also allows us to build our own net through a python script. Mininet can provide a connection between the switches in the net and the OpenFlow controller. Our purpose is to use the controller only to manage one of the two switches present in the net, in particular the switch in the subnet 10.0.1.0/24, as shown in the picture.
For running the net you must go in the mininet directory and type:
$ sudo python net.py
It will connect the controller by itself.
Each host in the subnet controlled by the OpenFlow controller has been configured with the default gateway 10.0.1.10 which corresponds to the IP address of the VIRTUAL_ROUTER.
Both routers involved in the protocol start sending UDP advertisements in broadcast to port 2020 with an interval of 1 second. The controller will elect as Master the router from which it received the ADV first. The other will become the Backup router.
The controller maintains the following data:
MASTER_ROUTER: represents the actual MAC and IP address of the elected Master router;LAST_MASTER_ADV: represents the last time that the controller has received an ADV from the Master;MASTER_DOWN_INTERVAL: represents the time interval to decleare Master no more active. Usually it is set accordingly to the interval time between two consecutive ADVs (1second in our case).
Every time the controller receives an ADV form the Master, it updates the LAST_MASTER_ADV.
If the controller receives an ADV from the Backup router and it is not receiving ADVs from the Master for a period greater than the MASTER_DOWN_INTERVAL, it will promote the Backup to Master. Moreover the controller will flood the new Master MAC address to all the hosts by means of a Gratuitous ARP Reply.
It is possible to run the protocol with a preemption mode in which it is possible to set a PRIMARY_ROUTER which, when active, must be the MASTER_ROUTER.
on_receive(source):
if MASTER_ROUTER is not set yet:
MASTER_ROUTER = source
LAST_MASTER_ADV = time.now
else if source is MASTER_ROUTER:
LAST_MASTER_ADV = time.now
else if LAST_MASTER_ADV - time.now > 1:
MASTER_ROUTER = source
LAST_MASTER_ADV = time.now
flood(MASTER_ROUTER_MAC)
else if preemption and source is PRIMARY_ROUTER and PRIMARY_ROUTER is not MASTER_ROUTER:
MASTER_ROUTER = source
LAST_MASTER_ADV = time.now
flood(MASTER_ROUTER_MAC)
Whenever the controller gets an ARP request for knowing which is the MAC address of the VIRTUAL_ROUTER, the controller will reply with the MAC address of the current MASTER_ROUTER.
The controller can be configured in order to adapt it for any network. The configuration is offered by means of REST resources which are exposed at http://127.0.0.1:8080/vrrm/config in a JSON format.
The default configuration is represented by the following json object:
{
"virtual_host_ip": "10.0.1.10"
"primary_router_ip": "10.0.1.1"
"preemption_mode": false
}
If you want to edit the above configurations you have to make a POST request to the resource http://127.0.0.1:8080/vrrm/config. For this purpose you can use the Firefox plugin called RESTED in which you build your own JSON payload and issue it.
In order to modify a specific parameter you have to put its name in the KEY field and its new value in the VALUE field.
In order to check the correcteness of the protocol just generate a data flow between the two networks using iperf tool. In mininet, open xterm for h11 and h21.
On h21 type:
$ iperf -u -s -i 1
and then on h11 type:
$ iperf -u -c 10.0.2.3 -i 1 -t 120
To simulate a failure of the link between S1 and R1 type:
mininet> link s1 r1 down
In the iperf server you should notice a decreasing of traffic for a small period of time and then, as soon as the new Master is elected, the flow restarts to increase.