Skip to content

Distributed Emulation

A topology can be partitioned across different machines using the FogbedDistributedExperiment class. This class enables centralized control of a cluster of remote machines.

Unlike the FogbedExperiment, now we need to add the virtual instances within Worker objects. These workers can be connected using tunnels through the Fogbed API to allow the comunication between distributed containers.

topology.py
from fogbed import (
    FogbedDistributedExperiment, 
    Container
)

exp   = FogbedDistributedExperiment()

cloud = exp.add_virtual_instance('cloud')
fog   = exp.add_virtual_instance('fog')
edge  = exp.add_virtual_instance('edge')

d1 = Container('d1')
d2 = Container('d2')
d3 = Container('d3')

exp.add_docker(d1, cloud)
exp.add_docker(d2, fog)
exp.add_docker(d3, edge)

worker1 = exp.add_worker(ip='192.168.0.151', port=5000)
worker2 = exp.add_worker(ip='192.168.0.152', port=5000)

worker1.add(cloud, reachable=True)

worker2.add(fog, reachable=True)
worker2.add(edge)
worker2.add_link(fog, edge, delay='10ms')

exp.add_tunnel(worker1, worker2)

try:
    exp.start()
    print(d1.cmd(f'ping -c 4 {d3.ip}'))

except Exception as ex: 
    print(ex)
finally:
    exp.stop()

In this example, 3 virtual instances are distributed among 2 workers. The cloud instance has been added to worker1, while the fog and edge were added to worker2.

Note

The reachable parameter establishes a connection between the virtual instance and a switch gateway present in each worker. Through this gateway, the virtual instances can be reached by others.

Subsequently, the workers are connected using the add_tunnel method.

Tip

Make sure the machines are connected and that the IP addresses and ports are configured according to the experiment script.

For the worker ip parameter, you can provide an IP address or hostname, and Fogbed will automatically resolve it.

To run that example, start a service in each worker with: 

sudo RunWorker -p=5000
copy and save the code to a file and run it with: 
sudo python3 topology.py

Setting Controller Address

By default, Fogbed runs an OpenFlow Controller on the machine that executes the experiment. You can manually run a controller and pass the address to the experiment. 

from fogbed import (
    FogbedDistributedExperiment
)

exp = FogbedDistributedExperiment(
    controller_ip='192.168.0.150',
    controller_port=6633
)

Tip

For the controller_ip parameter, you can provide an IP address or hostname, and Fogbed will automatically resolve it.

To start a controller, run the command controller -v ptcp:6633 on the target machine. You can also use other controllers, such as the POX Controller.

Limiting CPU and Memory

To limit CPU and memory within the workers, use the FogbedDistributedExperiment class to set the max_cpu and max_memory parameters based on the number of workers in the topology. 

from fogbed import FogbedDistributedExperiment

workers = 3
exp     = FogbedDistributedExperiment(max_cpu=workers * 0.5, max_memory=workers * 512)

  • max_cpu - sets the maximum percentage of CPU usage that can be consumed by the emulation.
  • max_memory - defines the maximum amount of memory in megabytes that the emulation can utilize.

Helper methods

The FogbedDistributedExperiment class offers some helper methods like:

add_docker (container: Container, datacenter: VirtualInstance)

  • Adds a container to a virtual instance.
  • Raises a ContainerAlreadyExists exception if name or ip already exists.

add_tunnel (worker1: Worker, worker2: Worker, **params: Any)

  • Adds a tunnel between two workers.

add_virtual_instance (name: str, resource_model: Optional[ResourceModel] = None) -> VirtualInstance

  • Creates a virtual instance.
  • Raises a VirtualInstanceAlreadyExists exception if name already exists.

add_worker (ip: str, port: int = 5000, controller: Optional[Controller] = None) -> Worker

  • Creates a worker.
  • Raises a WorkerAlreadyExists exception if ip already exists.

get_containers () -> List[Container]

  • Returns all containers of the emulation.

get_docker (name: str) -> Container

  • Returns a container by name.
  • Raises a ContainerNotFound exception if name doesn't exist.

get_virtual_instance (name: str) -> VirtualInstance

  • Returns a virtual instance by name.
  • Raises a VirtualInstanceNotFound exception if name doesn't exist.

get_virtual_instances () -> List[VirtualInstance]

  • Returns all virtual instances of the emulation.

get_worker (ip: str) -> Worker

  • Returns a worker by ip/hostname.
  • Raises a WorkerNotFound exception if ip/hostname doesn't exist.

remove_docker (name: str)

  • Removes a container by name.
  • Raises a ContainerNotFound exception if name doesn't exist.

start ()

  • Starts the experiment.

stop ()

  • Stops the experiment.