Skip to content

Experiments with measuring system energy consumption and performance on HEPScore Benchmark, and its relationship with CPU frequency

Notifications You must be signed in to change notification settings

andrii0yerko/ER-IPT-HACK_2023

Repository files navigation

ER IPT HACK 2023. Track 2: Calculations

Solution of the YAV-SSV team. Were the only team that reached the final.

The research is centered around measuring power consumption at different CPU frequencies and in different environments (local laptop, Cloud VM)

See TASK.md for task details, and report/presentation.pdf for a detailed overview of the given problem.

Details of the results and their analysis can be found in the report folder.

Various auxiliary shell scripts are used to set up the work environment. Notice that it is not complete, you may need to additionally install at least python3 and docker if your system does not have them.

Should be run from the root of the project, as further scripts expect to have some files (at least python env) to be present here.

Source code for the experiments run and their raw results.

Experiments can be grouped into two categories: local runs with different CPU frequencies and run on Azure.

Used tools: perf, powertop, hepscore benchmark with different workloads, docker

See subfolders for more details

Usage:

sudo experiments/experiment1/run_benchmarking_fort.sh

with no additional agruments

Script to process experiment logs and build plots

Usage

python analysis/visualize.py analysis/config.yaml --show

To interactively display plots

Or

python analysis/visualize.py analysis/config.yaml --out_dir report

To save plots and tables for the report

Adding data

Paths to experiments and their metadata should be written in analysis/config.yaml

Resulting tables, graphs, and their analysis.

About

Experiments with measuring system energy consumption and performance on HEPScore Benchmark, and its relationship with CPU frequency

Resources

Stars

Watchers

Forks

Packages

No packages published