Detect the computing platform

Oct 13, 2025 by pinkandtrash.com 74 views

Introduction

The WeatherGenerator is a powerful tool for generating weather data, but it requires adaptation to different High-Performance Computing (HPC) platforms. This article proposes a simple Python CLI application/script that can detect the computing platform, select the matching configuration, and customize the jobscript template and platform-specific WeatherGenerator configuration. This wrapper script will simplify the process of launching and configuring WeatherGenerator on various HPC platforms.

Problem Statement

The WeatherGenerator needs to be adaptable to different HPC computing platforms, which poses a challenge. Specifically, this requires:

Special filesystem paths
Jobscript submission customization

Solution

A simple Python CLI application/script will be created to detect the computing platform via the hostname, select the matching configuration, and use this configuration to customize arguments in the jobscript template and the platform-specific WeatherGenerator configuration. The configuration will be stored in JSON files named after the specific platform (e.g., juwels-cluster.json). These files will specify:

A platform unique string that can be used to match against the hostname
Platform-specific locations on the file system (e.g., those in #49)
Platform-specific Slurm arguments (e.g., partition, account, ...)

Implementation

The wrapper script will be implemented in Python, using the following steps:

Detect the computing platform: The script will use the hostname to detect the computing platform.
Select the matching configuration: The script will select the matching configuration from the JSON files based on the detected platform.
Customize the jobscript template: The script will use the selected configuration to customize the jobscript template.
Customize the platform-specific WeatherGenerator configuration: The script will use the selected configuration to customize the platform-specific WeatherGenerator configuration.

Configuration Files

The configuration files will be stored in JSON format, with each file named after the specific platform (e.g., juwels-cluster.json). Each file will contain the following information:

Platform unique string: A unique string that can be used to match against the hostname.
Filesystem locations: Platform-specific locations on the file system.
Slurm arguments: Platform-specific Slurm arguments (e.g., partition, account, ...).

Example Configuration File

Here is an example of what the juwels-cluster.json file might look like:

{
  "platform": "juwels-cluster",
  "filesystem_locations": {
    "input_data": "/path/to/input/data",
    "output_data": "/path/to/output/data"
  },
  "slurm_arguments": {
    "partition": "juwels",
    "account": "my_account"
  }
}

Alternatives

While YAML configuration files are more suitable for specifying complex workflows, JSON is more advantageous for this use case due to its simplicity.

Additional Context

This application could also be used to launch and configure a daemon for syncing artifacts to a remote model/artifact repository, as discussed in #45.

Implementation Options

This application can be implemented as a script embedded into the WeatherGenerator package and exposed via the pyproject.toml [projects.scripts] mechanism or as a separate package/project. The latter might be more suitable at a later point if the scope of the launch application grows too much.

Conclusion

In conclusion, a simple Python CLI application/script will be created to detect the computing platform, select the matching configuration, and customize the jobscript template and platform-specific WeatherGenerator configuration. This wrapper script will simplify the process of launching and configuring WeatherGenerator on various HPC platforms.

Future Work

Future work will include:

Implementing the wrapper script in Python
Creating the configuration files in JSON format
Testing the wrapper script on various HPC platforms

Organisation

Q: What is the purpose of the simple wrapper script?

A: The simple wrapper script is designed to detect the computing platform, select the matching configuration, and customize the jobscript template and platform-specific WeatherGenerator configuration. This will simplify the process of launching and configuring WeatherGenerator on various HPC platforms.

Q: How does the wrapper script detect the computing platform?

A: The wrapper script uses the hostname to detect the computing platform.

Q: What is the format of the configuration files?

A: The configuration files are stored in JSON format, with each file named after the specific platform (e.g., juwels-cluster.json).

Q: What information is stored in the configuration files?

A: The configuration files contain the following information:

Platform unique string: A unique string that can be used to match against the hostname.
Filesystem locations: Platform-specific locations on the file system.
Slurm arguments: Platform-specific Slurm arguments (e.g., partition, account, ...).

Q: Can I use YAML configuration files instead of JSON?

A: While YAML configuration files are more suitable for specifying complex workflows, JSON is more advantageous for this use case due to its simplicity.

Q: Can the wrapper script be used to launch and configure a daemon for syncing artifacts to a remote model/artifact repository?

A: Yes, the wrapper script can also be used to launch and configure a daemon for syncing artifacts to a remote model/artifact repository, as discussed in #45.

Q: Can the wrapper script be implemented as a separate package/project?

A: Yes, the wrapper script can be implemented as a separate package/project. This might be more suitable at a later point if the scope of the launch application grows too much.

Q: What is the benefit of using the wrapper script?

A: The wrapper script simplifies the process of launching and configuring WeatherGenerator on various HPC platforms, making it easier to use and manage.

Q: Who is responsible for implementing the wrapper script?

A: The implementation of the wrapper script is the responsibility of the JSC organisation.

Q: What is the next step after implementing the wrapper script?

A: After implementing the wrapper script, the next step will be to test it on various HPC platforms to ensure its functionality and accuracy.

Q: Can I contribute to the development of the wrapper script?

A: Yes, contributions to the development of the wrapper script are welcome. Please contact the JSC organisation to discuss how you can contribute.

Q: Where can I find more information about the wrapper script?

A: More information about the wrapper script can be found in the project documentation and on the JSC organisation's website.