Case Study: Self-adaptive Edge Computing System

Objective

To develop a self-sufficient edge computing system powered by PV, capable of dynamically adjusting its CPU cores and frequency based on available energy and user demand.

The system aims to meet user demand while remaining energy ≥ 0 for each time unit.

Input

• The energy available (time unit: hour)

• User demand (time unit: hour)

Goals:

• Meet User Demand: Ensure the system provides the required performance (computing resource the system can provide) to meet user demand (computing resource the user needs).
  • User demand is satisfied when performance ≥ user demand.
• Maximize Energy Efficiency: Optimize the use of energy consumption to prolong system operation and efficiency.
  • remaining energy should be ≥ 0 at the end of the simulation

Outcome

A dynamically self-adaptive edge computing system that efficiently manages its resources based on available energy and user demand for each time unit (hour), ensuring sustainable and efficient operation for autonomous vehicles requiring edge computing capabilities.

File Structure

/src
  |-- MOO4Modelica
  	|-- config.json
  	|-- config.py
  	|-- optimize_main.py
  	|-- parallel_computing.py
  	|-- optimization_libraries.py
  |-- OrchestrationWorkflow
      |-- ITSystem.mo  # Modelica model
      |-- energy_available_and_user_demand_data.txt  # input data
      |-- orchestration_config.json  # config file of the orchestration workflow
      |-- orchestrator.py
      |-- orchestration_wrapper.py
      |-- orchestration_configurator.py

Steps

Step 1: Set up the configuration in orchestration_config.json:

{
    "DATA_FILE_PATH": "data/energy_available_and_user_demand.txt",
    "CONFIG_PATH": "config.json",
    "MODEL_FILE": "ITSystem.mo",
    "MODEL_NAME": "ITSystem",
    "SIMULATION_TIME": 100,
    "TIME_CONFIG": {
        "START_TIME": 8,
        "END_TIME": 12,
        "TIME_UNIT": "hour"
    },
    "OBJECTIVES": [
        {"name": "remainingEnergy", "maximize": true},
        {"name": "performance", "maximize": true}  # computing power the system provides 
    ],
    "TUNABLE_PARAMETERS": {
        "PARAMETERS": ["activeCores", "cpuFrequency"],
        "PARAM_BOUNDS": {
            "activeCores": {
                "bounds": [1, 4],
                "type": "int"
            },
            "cpuFrequency": {
                "bounds": [1.0, 3.0],
                "type": "float"
            }
        }
    },
    "INPUT_PARAMETERS": {
        "available_energy": "availableEnergy",
        "user_demand": "userDemand"  # computing power the user needs
    },
    "CRITERIA": {
        "GOAL_EXPRESSION": [
            "evaluation_results['performance'] >= simulation_inputs['user_demand']",
            "evaluation_results['remainingEnergy'] >= 0"
        ]
    },
    "OPTIMIZATION_CONFIG": {
        "USE_SINGLE_OBJECTIVE": false,
        "ALGORITHM_NAME": "nsga2",
        "POP_SIZE": 5,
        "N_GEN": 2
    },
    "PLOT_CONFIG": {
        "PLOT_X": "",
        "PLOT_Y": "",
        "PLOT_TITLE": "",
        "ENABLE_PLOT": false
    },
    "LIBRARY_CONFIG": {
        "LOAD_LIBRARIES": false,
        "LIBRARIES": [
            {"name": "", "path": ""}
        ]
    },
    "N_JOBS": -1
}

Step 2: run python orchestrator.py:

Finally, you will see the final report:

Processing hour: 8  # 8 am
Parameters set: {'activeCores': 4, 'cpuFrequency': 3.00}
Performance: 930  # at 8 am user demand is high, the evaluated performance is 930 which satisfies user demand
User demand satisfied.

...

Processing hour: 12  # 12 am
Parameters set: {'activeCores': 2, 'cpuFrequency': 1.80}
Performance: 372  # at 12 am user demand is medium, the evaluated performance is 372 which satisfies user demand
User demand satisfied.

Final Report:
Hour 8: User demand satisfied with configuration {'activeCores': 4, 'cpuFrequency': 2.5}.

...

Hour 12: User demand satisfied with configuration {'activeCores': 2, 'cpuFrequency': 1.8}.

You can also visualize the final result:

At 8 AM, both goals are not satisfied.

At 9 AM, the first goal is not satisfied.

From 10 AM to 12 AM, both goals are satisfied.