engine

EnGINE Framework Brief Overview

The experiment campaigns are managed in four steps:

[ install ] --> [ setup ] --> [ scenario ] --> [ process ]

• install: install nodes with a image and prepare them to be available for experiments
• setup: do some generic setup tasks (install non included packages, etc.) so that nodes have all dependencies required for experiments
• scenario: run an experiment campaign -> multiple experiments around one topic
• process: do post-processing for a single experiment or over multiple experiments

For every step there is a playbook to execute it. All playbooks require one extra parameter that needs to be specified when executing the playbook: "pgroup" is the host group here all the commands should be applied to. The pgroup name points to a group inside the hosts file.

Playbook Execution

Playbooks are executed from the command line in the top-level inside the repository folder. The folder needs to be on a testbed management node in order to execute the "install" step / playbook.

• Run install: ansible-playbook plays/install.yml -e pgroup=nodes
• Run setup: ansible-playbook plays/setup.yml -e pgroup=nodes
• Run scenario: ansible-playbook plays/scenario.yml -e scene=<scenario-name>
• Run post-processing: ansible-playbook plays/process.yml -e scene=<scenario-name> -e scene_folder=<absolute-path-to-folder-with-scenario-results>
• Run all steps:ansible-playbook plays/all.yml -e scene=<scenario-name>

Extended by OMNeT++ Simulator

You can try the custom OMNeT++ project standalone. See README for the required OMNeT++ and INET framework versions.

The translation scripts and templates that generate an INI-file based on an EnGINE scenario definition via YAML files can be found in the Ansible role generate_ini. In subfolder files, run ./generate_ini.py -o <scenario-name>.ini -m mac <scenario-name>.

The playbook for running the simulation requires that appropriate versions of OMNeT++ and the INET framework, as well as the custom simulation project are installed at certain locations (see README). We use a playbook that does this installation (plus installation of any dependencies) from scratch at a selectable testbed node. See the playbook for details.

• Run setup: ansible-playbook plays/sim_setup.yml -e simulation_node=<node-name>

Once everything is setup, the actual simulation can be run by passing the respective scenario.

• Run scenario: ansible-playbook plays/sim_scenario.yml -e simulation_node=<node-name> -e scene=<scenario-name>

Processing of the (simulation) result files can be done manually via the script process.py

• Run result processing script: ./process.py -d ../../plays -f <folder-with-scenario-results> <scenario-name> -c -p -s
  • The -s flag is required for the results produced by the simulation playbooks. Without this flag the scripts can also be used to process results produced by the original engine playbooks.

Limitations of This Repository

To deploy EnGINE in your infrastructure, you need hardware managed from a central management host with SSH access to the individual nodes. If you have such deployments on hand, you can update the host_vars directory with your corresponding nodes.

For our deployments we are using pos [1]. To find out about the architecture, navigate to these publications:

• [1] Pos
• [2] EnGINE JNSM