Artifacts and results for the paper Accountable Carbon Footprints and Energy Profiling For Serverless Functions
These were captured and generated using the Ilúvatar serverless control plane. The main driver which captures energy and performance details are collected is in this set of files. For example it can pull data from Intel and AMD RAPL registers and query baseboard management controller IPMI data. These metrics are collected and fed into the FaasMeter system to determine per-function and per-invocation energy usage and carbon fotprints. Our control plane also possesses the capability to limit dispatches of enqueued invocations if the energy consumption of the hardware is too high.
To run the analysis and graphing scripts, the following dependencies are needed.
sudo apt install python3 python3-pipThen run this script which will generate all the plots.
./plot_all.shThese are all the plots from the paper that are generated by the artifacts and scripts in this repository. Traces can be found in the output-full-mc_a.json file in the corresponding linked directories.
Figure 2
Figure 6
Figure 8
Figure 10a
Figure 10b
Figure 10c
Figure 11
Figure 12
