Experiment 5: Mesh Networking

The fifth experiment adds a new class of nodes to the known AP and STA node: Mesh Points (MPs). In contrast to APs and STAs, MPs can neither be source nor sink of a traffic flow. In fact, MPs do not have an IP layer, but use a layer 2 path selection protocol to find the optimal path in the network between an AP and a STA or between two STAs. Therefore, the mesh network is completely transparent for the IP layer and all layers above.

In the next steps, the scenario in the following figure shall be simulated: A STA wants to communicate with an AP that is out of its communication range. Hence, it uses the intermediate MPs which provide service for association and transparent data forwarding between the AP and the STA. The distance between the nodes in the line can be varied; furthermore, the number of hops (and thus the number of MPs) can be changed.

Scenario of experiment 5

We recommend to create a new sub-campaign for experiment 5 and to copy the config.py and the campaignConfiguration.py from experiment 4 into the new simulation directory.

Creating mesh points

The transceiver configuration of a MP is the same as the configuration for an AP; hence, the same specialised configuration class from experiment 4 can be used; we choose to rename it to MyMeshTransceiver:

####################
# Node configuration

# configuration class for AP and MP BSS transceivers
class MyMeshTransceiver(wifimac.support.Transceiver.Mesh):
    def __init__(self, beaconDelay):
        super(MyMeshTransceiver, self).__init__(frequency = networkFrequency)

        # Transmission power
        self.txPower = dBm(20)

        # set the inital start delay of the beacon so that beacons from multiple APs do not collide
        self.layer2.beacon.delay = beaconDelay

        # rate adaptation strategy: Constant BPSK 1/2
        self.layer2.ra.raStrategy = Constant()

        # For frames above this threshold (in bit) RTS/CTS will be used
        self.layer2.rtsctsThreshold = 8e6

Then, the MPs can be created using a for-loop; the only difference to the creation of an AP is the usage of the createMP function of the node creator:

for i in xrange(numHops-1):
    mpConfig = wifimac.support.Node(position = openwns.geometry.position.Position((i+1)*distance,0,0))
    mpConfig.transceivers.append(MyMeshTransceiver(beaconDelay = 0.001+(i+1)*0.001))
    mp = nc.createMP(idGen, managerPool, mpConfig)
    mp.logger.level = commonLoggerLevel
    mp.dll.logger.level = dllLoggerLevel
    WNS.simulationModel.nodes.append(mp)
    mpIDs.append(mp.id)
    mpAdrs.extend(mp.dll.addresses)
    print "Created MP at (", (i+1)*distance, ",0,0)",
    print "with id ", mp.id, " and addresses ", mp.dll.addresses

Please note that it is ensured that beacons are transmitted by the MPs at different points in time so that collisions of beacons with other beacons are impossible.

Remaining scenario configuration

The rest of the scenario configuration resembles the setting of the experiment 1: Only one AP and one STA has to be created; this time, the specialised configuration classes should be used for convenience.

Other changes to the configuration file are:

• One new parameter is required: numHops
• The sizeX of the scenario has to be enlarged, depending on the distance and the number of hops.
• The number of nodes, given as parameter to the virtual pathselection server, has to be adapted to the number of MPs.
• The evaluation of the wifimac requires as a parameter the number of hops.

Experiments

For the experiments, the campaignConfiguration of experiment 2 can be used; only the new parameter numHops has to be added. In the following, the nodes shall be separated by a distance so that the neighbours can communicate with each other, but the neighbor’s neighbors cannot. Hence, the 2-hop neighborhood shall be hidden nodes.

1. Determine the saturation throughput of the string with at least 2 hops if only unidirectional traffic (e.g. only downlink) is used.
2. Mesh networks often have the problem that the “boundary” nodes have much less neighbors as the central nodes; Hence, they see the channel more often as idle and thus “overflow” the central nodes. To simulate this effect, determine the saturation throughput of the string with bidirectional traffic, e.g. 50% uplink and 50% downlink traffic. What percentage of the unidirectional traffic can be reached? How can the difference be explained?

Note

An example config.py and campaignConfiguration.py can be found in myOpenWNS/tests/system/wifimac-tests/PyConfig/experiment5.