The GPU Dashboard is an application for displaying resource utilization information about a GPU enabled server.
The GPU Dashboard enables you to view the following information:
- GPU Memory
- What GPUs are in use
- What users are using the GPUs
- How much memory is each GPU process consuming
- GPU Storage
- View CPU load by GPU process
- Storage
- View disk storage by directory name
- How much storage are each of the largest directories consuming.
CPU Load
Storage
This application is composed of three main parts:
The logger runs on the GPU server and contains a set of scripts that run periodically using a cron job to post GPU and storage data to the server.
The server component listens for incoming requests to either post new logging information from the GPU or to provide data to the client.
The client is the dashboard front end which requests data from the server and displays the data in an easy to user visual form.
Upload the logger files to the GPU server and add a cron job to periodically run the following scripts to report data to the web server:
python3 report-gpus.py
python3 report-storage.py
This web application is Dockerized so you should be able to run it by simply uploading this project to your web server and running:
docker compose up
Distributed under the MIT License. See LICENSE for more information.
Abe Megahed - (mailto:amegahed@wisc.edu) - email