Last Updated: 10/12/2022

Model Onboarding

Thank you for your interest in delivering your model to CCMC as a service to the community! Model developers, please read our Rules of the Road [PDF] on code acceptance and access.

Model Onboarding process flow

Right-click and open in new tab to view larger image.

There are 5 main stages of Model on-boarding process:

  1. Pre-onboarding

    If your project is at the proposal stage, we ask you to fill out a short pre-onboarding questionnaire.

    A note on computational resources: CCMC will generally aim to meet all computational and technology resource requirements of the proposed model for Runs-on-Request. If we can't meet the requirements with our in-house infrastructure, we can use resources at NASA HECC to execute the Runs-on-Request. We are open to all proposed innovative ways that CCMC can implement the proposed project.

  2. Preparation

    Modeler contacts CCMC to receive and complete pre-installation questionnaire (Word or PDF). A CCMC point-of-contact is assigned. The questionnaire information is used to add the model to the CCMC Metadata Registry (CMR) which automatically populate CCMC Model Catalog.

    The modeler also provides sample model output data, visualizations (possibly with data reader or visualization source code) and identifies how to prepare inputs and run the model.

    CCMC confirms the model metadata in CMR, identifies computing resources, creates a model visualization sample page (similar to this page) and a simulation output page.

  3. Model Implementation

    All model onboarding and initial testing at CCMC is now done collaboratively with the model developer, on the cloud. We call this the CCMC/Model Developers collaborative cloud environment. For this purpose CCMC will provide account access to a cloud-based server, including a model-specific account that only CCMC and the model developers have access to. CCMC uses the Amazon Web Services (AWS) Commercial Cloud. These cloud services meet the security criteria of Federal Information Security Management Act (FISMA) Low standards, and have NASA Authority to Operate (ATO).

    For certain projects, we may also work collaboratively in a shared space on NASA HECC supercomputers.

    Modeler supplies model source code and documentation (makefiles, preprocessing, run and postprocessing scripts, user manual), and begins the installation process on the provided cloud instance. The model developer and CCMC iterate and perform test simulations together.

    CCMC creates internal documentation to track the model onboarding process, source controls the code and documentation, re-installs the model on on-site servers at GSFC and performs tests. CCMC also creates a prototype model output and web visualization page for a test simulation.

  4. Test/Evaluation

    CCMC creates a prototype run submission web interface for Instant Runs (IR) or Runs-on-Request (RoR), for review by the modeler. CCMC and the modeler test the pipeline from end-to-end (executes model run, post-processing, and publishing results to the web).

    CCMC recommends that the model development updates are installed and tested periodically on the collaborative cloud environment and/or the shared space on NASA HECC, once the proposed project begins. The model developers and CCMC are expected to continually document the installation and testing process.

  5. Production and Upgrades

    Once the modeler is satisfied with the results, CCMC offers the model to the research community. CCMC will continue work with modeler to address any bug fixes and visualization issue(s) as issues occur. Model upgrades can be provided at any time and these are onboarded in the (4) Test/Evaluation step.

Close collaboration as well as strong commitment are needed from both the model developer(s) and CCMC in order to ensure any given model installation is successful and useful for the community. As such CCMC highly encourages and recommends model developer(s) to get in touch with CCMC as soon as possible, if they are interested in hosting their model(s) at CCMC. The model developer(s) can then work with the assigned CCMC POC to understand the detailed steps and expectation of each stage of the ‘model on-boarding' process.

If you would like to initiate the process or have any questions, please contact CCMC.