Running the Experiment

In this demo, we are going to investigate the effect of the location of a datacenter on its carbon footprint. The same data center is placed in four different locations; Belgium, Germany, Netherlands, and France. Because they are getting their energy from different sources, their carbon emission is different. This demo is based on the work done for the workshop paper FootPrinter: Quantifying Data Center Carbon Footprint which can found here.

info

In this tutorial, we will learn how to create and execute a simple experiment in OpenDC.

This is tutorial is based on a "Carbon emission demo" which can be found on the OpenDC github.

Download the demo [here] to run in an interactive notebook.

Running this demo requires OpenDC. Download the latest release here and put it in this folder.

1. Carbon Footprint

With the rise of high-computational workloads, such as Artificial Intelligence, datacenters are using a lot of energy. Because of this datacenters have become increasingly large contributors to the global footprint. A common definition of the carbon footprint of a datacenter is its total carbon emissions. In this demo, we are looking at the operational carbon emission of a datacenter. Operational carbon is all carbon that is emitted when producing the energy required to run a workload.

Carbon Intensity

Determining the operational carbon emission of a datacenter requires more than just the energy used. Equally important is the source of the energy. Using energy from non-green sources, such as coal, can emit up to 20x times as much carbon as green energy, such as solar. Carbon Intensity defines the amount of carbon emitted per unit of energy. Multiplying the energy usage of the energy by the carbon intensity of the energy source determines the carbon emission.

Green energy is primarily gained from natural phenomena, such as wind or sunlight. This results in a continuously changing mix of available energy. This means the carbon intensity, even at the same location, is rarely stable over a long period of time. Below, we see the carbon intensity of the Dutch energy grid during a period of a week:

Carbon Intenisity Figure taken from FootPrinter

Due to the fluctuating availability of green energy (the Netherlands relies on solar and wind) there is a large variation in carbon intensity over time. This shows that to reduce the carbon emissions it is not just important to reduce energy usage, but also when and where tasks are executed.

2. Carbon Traces

OpenDC can determine the carbon emission of a datacenter when provided a carbon trace. Carbon traces define the carbon intensity of the energy in a specific region during a particular time period.
Historical carbon traces can be collected from services such as entsoe-e or electricitymaps.

Carbon traces are saved as parquet files. Lets see how they look.

Input

import pandas as pd

df_carbon = pd.read_parquet("carbon_traces/BE_2021-2024.parquet")

Output

Input

df_carbon.head()

Output

	timestamp	carbon_intensity
0	2021-01-01 00:00:00	254.44
1	2021-01-01 01:00:00	247.21
2	2021-01-01 02:00:00	219.78
3	2021-01-01 03:00:00	191.87
4	2021-01-01 04:00:00	211.01

A carbon trace contains a large number of samples containing a timestamp and carbon intensity. During simulation, OpenDC matches the simulation time with the closest sample time.

Note: OpenDC will always try to match to a sample. If the simulation exceeds the range of the carbon trace, it will select the first or last sample.

3. Defining Topologies

Carbon traces are activated by the user in the topology file in the powerSource variable. The following is a topology that is connected to the Belgium energy grid:

{
    "clusters": [
        {
            "name": "C01",
            "hosts": [
                {
                    "name": "H01",
                    "cpu": {
                        "coreCount": 16,
                        "coreSpeed": 2100
                    },
                    "memory": {
                        "memorySize": 100000
                    },
                    "powerModel": {
                        "modelType": "sqrt",
                        "power": 400.0,
                        "idlePower": 32.0,
                        "maxPower": 180.0
                    },
                    "count": 277
                }
            ],
            "powerSource": {
                "carbonTracePath": "carbon_traces/BE_2021-2024.parquet"
            }
        }
    ]
}

First we see that the datacenter has 277 nodes containing a 16 core CPU. Next, we see that the carbonTracePath of the powerSource is pointed to the Belgium carbon trace.

Similarly, we can create topologies for the three other contries. See the topologies of Germany, France, and Netherlands

4. Executing the experiment

In this tutorial we want to investigate the impact of the location of a data center on its carbon footprint. We have defined four topologies that are connected to different carbon traces. We want to run the same workload using the different topologies. This can be done using the following experiment file:

{
    "name": "carbon",
    "topologies": [
        {
            "pathToFile": "topologies/surfsara_BE.json"
        },
        {
            "pathToFile": "topologies/surfsara_DE.json"
        },        
        {
            "pathToFile": "topologies/surfsara_FR.json"
        },        
        {
            "pathToFile": "topologies/surfsara_NL.json"
        }
    ],
    "workloads": [
        {
            "pathToFile": "workload_traces/surf_month",
            "type": "ComputeWorkload"
        }
    ],
    "exportModels": [
        {
            "exportInterval": 3600,
            "printFrequency": 168,
            "filesToExport": [
                "host",
                "powerSource",
                "service",
                "task"
            ]
        }
    ]
}

Running this experiment, will run four simulations, one for each topologies. All four simulations will run a month long workload trace based on the surf LISA computer.

The experiment can be executed directly from the terminal. While running the experiment, OpenDC periodically prints information about the status of the simulation. In this experiment, OpenDC prints every week, but this can be changes using the exportModel.

Input

import subprocess

pathToScenario = "experiments/carbon_experiment.json"
subprocess.run(["OpenDCExperimentRunner/bin/OpenDCExperimentRunner", "--experiment-path", pathToScenario])

Output

[32m================================================================================[0m [34m Running scenario: 0 [0m [32m================================================================================[0m [34m Starting seed: 0 [0m

Simulating... 0% [ ] 0/4 (0:00:00 / ?) 12:18:37.303 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 168 hours: Tasks Total: 21607 Tasks Active: 124 Tasks Pending: 0 Tasks Completed: 21483 Tasks Terminated: 0

12:18:38.659 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 336 hours: Tasks Total: 30276 Tasks Active: 114 Tasks Pending: 0 Tasks Completed: 30162 Tasks Terminated: 0

12:18:40.212 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 504 hours: Tasks Total: 58339 Tasks Active: 145 Tasks Pending: 0 Tasks Completed: 58194 Tasks Terminated: 0

12:18:41.502 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 672 hours: Tasks Total: 66103 Tasks Active: 228 Tasks Pending: 0 Tasks Completed: 65875 Tasks Terminated: 0

[32m================================================================================[0m [34m Running scenario: 1 [0m [32m================================================================================[0m [34m Starting seed: 0 [0m

Simulating... 25% [======== ] 1/4 (0:00:10 / 0:00:30) 12:18:45.457 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 168 hours: Tasks Total: 21607 Tasks Active: 124 Tasks Pending: 0 Tasks Completed: 21483 Tasks Terminated: 0

12:18:46.375 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 336 hours: Tasks Total: 30276 Tasks Active: 114 Tasks Pending: 0 Tasks Completed: 30162 Tasks Terminated: 0

12:18:47.487 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 504 hours: Tasks Total: 58339 Tasks Active: 145 Tasks Pending: 0 Tasks Completed: 58194 Tasks Terminated: 0

12:18:48.382 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 672 hours: Tasks Total: 66103 Tasks Active: 228 Tasks Pending: 0 Tasks Completed: 65875 Tasks Terminated: 0

[32m================================================================================[0m [34m Running scenario: 2 [0m [32m================================================================================[0m [34m Starting seed: 0 [0m

Simulating... 50% [================ ] 2/4 (0:00:17 / 0:00:17) 12:18:51.612 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 168 hours: Tasks Total: 21607 Tasks Active: 124 Tasks Pending: 0 Tasks Completed: 21483 Tasks Terminated: 0

12:18:52.377 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 336 hours: Tasks Total: 30276 Tasks Active: 114 Tasks Pending: 0 Tasks Completed: 30162 Tasks Terminated: 0

12:18:53.505 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 504 hours: Tasks Total: 58339 Tasks Active: 145 Tasks Pending: 0 Tasks Completed: 58194 Tasks Terminated: 0

12:18:54.466 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 672 hours: Tasks Total: 66103 Tasks Active: 228 Tasks Pending: 0 Tasks Completed: 65875 Tasks Terminated: 0

[32m================================================================================[0m [34m Running scenario: 3 [0m [32m================================================================================[0m [34m Starting seed: 0 [0m

Simulating... 75% [======================== ] 3/4 (0:00:23 / 0:00:07) 12:18:57.573 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 168 hours: Tasks Total: 21607 Tasks Active: 124 Tasks Pending: 0 Tasks Completed: 21483 Tasks Terminated: 0

12:18:58.432 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 336 hours: Tasks Total: 30276 Tasks Active: 114 Tasks Pending: 0 Tasks Completed: 30162 Tasks Terminated: 0

12:18:59.582 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 504 hours: Tasks Total: 58339 Tasks Active: 145 Tasks Pending: 0 Tasks Completed: 58194 Tasks Terminated: 0

12:19:00.543 [[33mWARN [m] org.opendc.compute.simulator.telemetry.ComputeMetricReader - Metrics after 672 hours: Tasks Total: 66103 Tasks Active: 228 Tasks Pending: 0 Tasks Completed: 65875 Tasks Terminated: 0

Simulating... 100% [=================================] 4/4 (0:00:28 / 0:00:00)

CompletedProcess(args=['OpenDCExperimentRunner/bin/OpenDCExperimentRunner', '--experiment-path', 'experiments/carbon_experiment.json'], returncode=0)

Running the simulation has created the output folder containing information about the experiment. In the next part we will use these files for analysis and vizualization.

Running the Experiment

1. Carbon Footprint​

Carbon Intensity​

2. Carbon Traces​

Input​

Output​

Input​

Output​

3. Defining Topologies​

4. Executing the experiment​

Input​

Output​

1. Carbon Footprint

Carbon Intensity

2. Carbon Traces

Input

Output

Input

Output

3. Defining Topologies

4. Executing the experiment

Input

Output