Offloading and Scheduling of Real-Time Tasks on Edge Resources in IIoT Environments

To build and run this project make and the Rust toolchain are required. Detailed installation instructions can be found at https://www.rust-lang.org/tools/install

Building the Project

Clone the repository.

Navigate into the source code directory:

cd ./code/

Build the project:

cargo build --workspace
cargo build -r

Build the dummy task:

cd ../tasks/
make
cd ../code/

Running Manually

Running the scheduler, worker and clients:

Scheduler

The scheduler needs to be started first.
From the ./code/ directory navigate to the scheduler directory:

cd scheduler

Starting the scheduler:

RUST_LOG=<log_level> cargo run <port> <allocation_strategy> <delay_factor>

The log_level can be one of trace, debug, info, warn, error, fatal.
The port can be chosen freely.
The allocation_strategy can be one of GS, PSFF, PSBF, PSWF. PSWF is the recommended option (stands for partitioned schedule with worst fit allocation).
The delay_factor can be a decimal number greater or equal to 0.0. Recommended is a value between 1.0 and 5.0.

Worker

Open another terminal. Repeat the following steps for as many workers as you would like to start.

From the ./code/ directory navigate to the worker directory:

cd worker

Starting the worker:

RUST_LOG=<log_level> cargo run <scheduler_ip:port> ../../tasks/target/task

The log_level can be one of trace, debug, info, warn, error, fatal. The events for the evaluation metrics are logged as trace events.
The scheduler_ip:port needs to be in the format 123.123.123.123:12345.

Client

Open another terminal. Repeat the following steps for as many clients as you would like to start.

From the ./code/ directory navigate to the client directory:

cd client

Starting the client:

RUST_LOG=<log_level> cargo run <scheduler_ip:port> <execution_time> <laxity> <return_data_bytes> <interval_lambda> <rng_seed>

The log_level can be one of trace, debug, info, warn, error, fatal. The events for the evaluation metrics are logged as trace events.
The scheduler_ip:port needs to be in the format 123.123.123.123:12345.

For the task parameters please refer to the thesis paper. An example would be:

RUST_LOG=trace cargo run 192.168.178.78:1337 1000 100 1 0.5 0

The client will now start to submit tasks to the scheduler.

Everything is logged to stderr by default.

Running the Evaluation Scenarios

The evaluation scenarios need Mininet to be installed. Sudo rights might be necessary. Detailed instructions can be found at https://mininet.org/download/.
The analysis dependencies are installed with Poetry (https://python-poetry.org/docs/).

From the root of the repository navigate to the eval directory:

cd eval

Install the required Python dependencies:

poetry install

Run one of the scenarios:

sudo python test_scenarios/scenario3b.py

Depending on the Mininet installation, sudo might not be necessary.

For details about the parameters please refer to the respective section in the paper and the scenario scripts.

All logs for an executed scenario are put into the respective test_scenarios/scenario<...>_logs/ directory.

Analysis

Two analysis scripts are available for analysing the output logs of the scenarios. Pass the respective log directory as parameter:

poetry run python analysis.py test_scenarios/scenario<...>_logs/
poetry run python scatter.py test_scenarios/scenario<...>_logs/