Tools for running research benchmarks on shared HPC systems. If you just cloned this directory try running the following command to get started:
source ./HPCC_INSTALL.sh
More detailed instructinos are below.
Assuming the above installation worked, there are multiple benchmarks that can be run from this directory. Look for the '''<>_Batch.sb''' file to run each exmple. Supporting files will have other names and extentions including:
- '''<>_Results.ipynb''' - Jupyter notebooks to view results form the benchmark.
- '''<>_Results.sb''' - File to consolodate data from the benchmarks. (only required on the really big runs)
- '''<>_<>.py''' - Python file used by the other scripts to manage the run.
For these steps to work you need access to the "beta" group to use a singularity image in jupyter. If you have access to this group you should be able to run this singularity image using the following settings:
- Go to http://ondemand.hpcc.msu.edu and log in using your MSU NetID / Password
- Click on "Interactive Apps" and select "Jupyter (beta)"
- Find the section labeled "Jupyter Location" and select "Singularity" from the dropdown menu.
- Use all of the default settings except change the "Path to Overlay Image" to the
overlay.img
file in this folder.
In case you need it, here are resonable useful settings:
Setting | Default Values |
---|---|
Number of Hours | 4 |
Number of cores per task | 1 |
Amount of Memory | 5gb |
Jupyter Location | Singularity |
Jupyter Path | /miniconda3/bin |
Path to Singularity Image | /opt/software/CentOS.container/7.4/bin/centos |
ReadOnly | unchecked |
Path to Overlay Image | Set this to your overlay.img path |
- Open the ContinuousArray.ipynb and run to view results.