The Fog Development Kit (FDK)

Developed by: SCU Internet of Things Research Lab (SIOTLAB)

The Fog Development Kit is a comprehensive fog application development platform, intended to simplify and lower the cost of creating fog applications.

Terms: OpenDaylight = ODL, Open vSwitch = OVS

INSTALLATION

Installation is a multi-step process. The Fog Development Kit simplifies the process of developing fog applications on a network. As such, the construction and setup of an entire network is required.

You must use:

• OpenFlow switches (with OpenFlow 1.3 + OVSDB support - We recommend OVS)
• A Controller (Runs the FDK and ODL)
• Fog nodes
• Edge devices

OpenFlow Switch Config

• It is assumed you can handle the configuration of the network and the OpenFlow Switches.
• For OVS users: You must specify the controller machine as both the controller AND manager of every switch bridge in the topology.
• We provide scripts in ./vm_config/ovs/ as a template for setting up OVS machines, but you must modify them according to your needs. See ./vm_config/ovs/README.md for more info on this sample configuration process.

Controller Config

• Download the FDK onto the machine
• install ODL 0.8.4 and Python3
• In ODL, run: feature:install odl-l2switch-all odl-restconf-all features-openflowplugin odl-yangtools-common odl-ovsdb-utils odl-l2switch-switch-ui odl-mdsal-all odl-ovsdb-southbound-impl-ui odl-l2switch-switch-rest
• In the FDK folder, modify the controller IP address in fdk_conf.json.
• Install the following Python3 packages in your python environment for FDK: requests, docker
• Install Docker via the Docker docs: https://docs.docker.com/install/
• Ensure Docker can run in non-root mode (instructions in the Docker docs)
• After ODL starts, the FDK can be started by running python3 fdk.py

Fog Node Config

• Fog nodes must greet the controller, which identify themselves as fog nodes and enabling services to be instantiated on them. See vm_config/fog/fog.py for a sample greeting script. This script will also attempt to begin reporting resource utilization to the FDK.
• Fog nodes must also install Docker and configure it to be able to be managed by the controller. See vm_config/fog/setup.sh (modify to your needs) and vm_config/fog/notes.txt for more details. This step enables remote instantiation of containers on the devicevia Docker Swarm.
• Fog nodes must have Docker images built, which can be instantiated as containers and requested by end-devices with a specific amount of resources. Fog nodes can only have containers instantiated on them after greeting the FDK. Docker images are built by first creating a Dockerfile and then running the docker build command with the desired parameters. A sample Dockerfile is provided in dockerfiles/fog_services/iperf_app/Dockerfile. You must give a name (or repository, in Docker terms) to the image, which is then requested by a service request from an end-device

End Device Config

• You must setup your end-devices to greet the controller. After a successful greeting, it can then run applications. A sample greeting python script can be found in dockerfiles/host_nodes/greeting.py
• A sample application script can be found in iperf_app.py.

APPLICATION CREATION

• This details how you can build a basic application, using the provided templates

• Create a sample fog node Docker service and end-device application.

• A sample fog node Docker service can be found in dockerfiles/fog_services/iperf_app. Run docker build with the appropriate parameters to build a Docker image This image name/repository can then be requested by service requests, which will instantiate the desired image as a container in the fog. Note: Please refer to the Dockerfile if you encounter any issues. Typically you must clone fog-development-kit into your $HOME directory and then build the Docker image from that directory. However, you can change the Dockerfile accordingly if you don't want to do that.

• A sample end-device application script can be found in test/edge_requests/edge_request_iperf.py. This application:

  • Creates and issues a service request. Parameters for resources and the image/repository name are taken at the command line.
  • If the previously-built iperf_app_fog image is requested (again, this is entered at the command line), an Iperf3 server starts within a containerized service in the fog.
  • Upon success, starts an Iperf3 client that connects to the Fog node IP and port returned in the success response by the FDK.
  • Finally, upon termination or Ctrl-C, the end-device issues a shutdown request, completely terminating the service and de-allocating any associated resources.

• Object detection application:

  • end-device application is within test/edge_requests/edge_request_quic_client.py and test/quic_client. Execute the application from within test/edge_requests.
  • fog-device service is within dockerfiles/fog_services/quic_server. Please see dockerfiles/fog_services/quic_server/README.md for setup information

EXECUTION SAMPLE

Start ODL, and let it load completely. Then start the FDK, the greeting applications, and the fog greeting applications in any order. Once an end-device confirms it's greeting has been processed (It will receive an empty ACK message from the FDK upon processing), it can begin issuing service requests and running fog applications.

Then run applications on end-devices that issue service requests and then communicate with that service in the fog. Each response contains a code indicating FAILURE or SUCCESS. If SUCCESS is detected, then the desired service is instantiated in the fog. Contained within the response will be an "ip" and "port" field, which can be used to then communicate with the instantiated service. Once complete, the application should issue a shutdown request to deallocate the service.

Upon turning off the system, hit Ctrl-C once to begin shutdown in the FDK. Wait until the message "IT IS SAFE TO EXIT" is displayed before completely terminating the program. This will ensure that flows, packet queues, etc. are cleaned up and remove in their entirety.