Replicating results

Prerequisite: Follow the instructions on the Installation and Environment Setup Page. Download the dataset folder from google drive or Zenodo and save it in the repo folder.

Experimental setup

To run the experiments we used a setup with the fog server in London, UK. We had 4 Azure B2s and 2 Azure B4ms in Azure London UK datacenter and 2 Azure B4ms and 2 Azure B8ms virtual machines in Azure East US datacenter.

Evaluation Metrics

We use the following metrics for evaluation: Energy consumption Response time SLA Violations Fairness Migration Time Scheduling Time Wait time

Generating your own results

To generate your own results run the following command with different schedulers.

$ python main.py -e VLAN -m 0

Save the pickle file generated in logs viz logs/*/*.pk to all_datasets/framework/$Algo/*.pk, where $Algo is the name of the scheduling algorithm which is one of A3C, GA, GOBI, GOBI2, LR-MMT, MAD-MC, and POND.

Similarly, for simulator run the following with different schedulers:

$ python main.py

Save the pickle file generated in logs viz logs/*/*.pk to all_datasets/simulator/$Algo/*.pk, where $Algo is the name of the scheduling algorithm which is one of A3C, GA, GOBI, GOBI2, LR-MMT, MAD-MC, and POND.

Generating graphs

To generate graphs make sure you have all_datasets folder in the repo. Run the following command:

$ python grapher.py framework
$ python grapher.py simulator

All graphs would be stored in the results/ folder.