Docker version of PowerAPI

Lightweight docker images of PowerAPI packaged as docker images.
The PowerAPI configuration directory is located at /powerapi/conf for all images.
Please refer to the Wiki for more information about PowerAPI configuration.

PowerAPI: CLI

This image is used to run a various choices of software-defined power meters.

Usage

The minimal command line to run PowerAPI (print the help message):

docker run --rm --privileged spirals/powerapi-cli

The --privileged flag gives all capabilities to the PowerAPI container. (required to get metrics about system usage for the various modules)

Example

Monitoring the machine using a PowerSpy power meter

Content of the powerapi-cli.conf, adapt the MAC address accordingly:

powerspy.mac = "00:0B:CE:XX:XX:XX"

Command to run to start the monitoring using the PowerSpy power meter:

docker run \
--rm --privileged --net=host \
-v $PWD/powerapi-cli.conf:/powerapi/conf/powerapi-cli.conf \
spirals/powerapi-cli \
modules powerspy monitor --frequency 1000 --console

The --net=host option is mandatory to be able to access the host bluetooth card inside a container.
If you have problem to connect to the PowerSpy, please check your ability to pair with PowerSpy from the host machine using bluetoothctl.

Monitoring a process using the ProcFS module (Linux only)

Content of the powerapi-cli.conf, adapt the CPU TDP accordingly:

powerapi.cpu.tdp = 130

Command to run to start the monitoring of the firefox application and the 1234 PID using the ProcFS module:

docker run \
--rm --privileged --pid=host \
-v $PWD/powerapi-cli.conf:/powerapi/conf/powerapi-cli.conf \
spirals/powerapi-cli \
modules procfs-cpu-simple monitor --frequency 1000 --apps firefox --pids 1234 --console

The --pid=host flag is required to be able to monitor the processes running on the host machine.

Monitoring a Docker container using the ProcFS module (Linux only)

Content of the powerapi-cli.conf, adapt the CPU TDP accordingly:

powerapi.cpu.tdp = 130

Command to run to start the monitoring of the stress container using the ProcFS module:

docker run \
--rm --privileged --net=host --pid=host \
-v /sys:/sys -v /var/run/docker.sock:/var/run/docker.sock \
-v $PWD/powerapi-cli.conf:/powerapi/conf/powerapi-cli.conf \
spirals/powerapi-cli \
modules procfs-cpu-simple monitor --frequency 1000 --containers stress --console

Access of the hosts /sys is required to access the Docker containers cgroups inside the container.
The Docker socket is used to resolve the name, short and full ID of the container and check its existence.

PowerAPI: Sampling

This image is used to build a CPU power model used by the libpfm module of PowerAPI.

Usage

First you have to configure PowerAPI, create a powerapi-sampling-cpu.conf file and adapt accordingly the following settings:

powerspy.mac = "00:0B:CE:XX:XX:XX"

powerapi.cpu.topology = [
  { core = 0, indexes = [0, 4] }
  { core = 1, indexes = [1, 5] }
  { core = 2, indexes = [2, 6] }
  { core = 3, indexes = [3, 7] }
]

powerapi.sampling.dvfs = true
powerapi.sampling.turbo = true

powerapi.cycles-polynom-regression.cpu-base-frequency = 0.133
powerapi.cycles-polynom-regression.cpu-max-frequency = 2.66
powerapi.cycles-polynom-regression.unhalted-cycles-event = "CPU_CLK_UNHALTED:THREAD_P"
powerapi.cycles-polynom-regression.ref-cycles-event = "CPU_CLK_UNHALTED:REF_P"

interval = 1s
powerapi.actors.timeout = 15s
powerapi.sampling.interval = ${interval}
powerspy.interval = ${interval}
powerapi.sampling.steps = [100, 25]
powerapi.sampling.step-duration = 10

Launch the CPU sampling with the following command:

docker run \
--rm --privileged --net=host --pid=host \
-v $PWD/powerapi-sampling-cpu.conf:/powerapi/conf/powerapi-sampling-cpu.conf \
-v $PWD/powerapi-sampling-cpu-results:/powerapi/results \
spirals/powerapi-sampling-cpu

The results of the sampling will be in the powerapi-sampling-cpu-results folder of the host.
The CPU power model will be located at the powerapi-sampling-cpu-results/computing/libpfm-formula.conf path.