Domino 5.0: Up-Level Your 'Laptop-in-the-Cloud' Experience for the Best of Both Worlds

Introducing Durable Workspaces with Checkpoints in Domino 5.0

To accelerate data science engagement and productivity, reproducibility is essential. It not only boosts exploration, but also results in increased model velocity by building upon a proven model or providing the ability to co-develop a model with others.

Durable Workspaces, which were introduced in Domino 4.4, are persistent development environments that you can start, stop, and restart. Each workspace has a persistent volume that houses your files, and your changes are kept from one workspace session to the next so that you can decide when to commit changes to version control.

In Domino 5.0, a ‘checkpoint’ is now created every time a user syncs within a workspace. This takes our ‘laptop on cloud’ experience to the next level. Now with these checkpoints, you can browse past work, restore, and branch experiments in new directions or foster collaboration within an organization.

Benefits:

• Create a workspace session from a desired point by viewing sessions and commit history.
• Easily review workspace session history along with commits to recreate work from a specific checkpoint.
• Quickly recreate workspaces from published models to be able to debug a production issue and rapidly redeploy a model.

How it Works

NOTE: see the Domino documentation for some prerequisites for Durable Workspaces with Checkpoints.

Revisiting and Starting from a Previous State of a Workspace

In addition to workspace sessions, you can view checkpoints for each session. Checkpoints are commits that you can return to at any time in order to review the history of your work, and/or branch your work in new directions. Checkpoints are created every time you synchronize changes to artifacts or code within a workspace.

You can preview the artifacts or code from any commit to identify the checkpoint from which you want to recreate a workspace.

When you recreate a workspace from a previous commit, Domino creates a new branch where you can perform new model development or training. Recreated workspaces fundamentally behave similar to existing workspaces. You can find detailed information about this workflow here.

Remediate Models that are Drifting or Decaying

The new integrated model monitoring capabilities in Domino 5.0 can automatically identify problems with data drift and model quality. You can use Automated Insights to explore the reasons and determine whether model remediation is needed. When it is, you can recreate the workspace that was used to deploy the model so you can update the model code or retrain it with the latest production data. The workflow is similar to the workflow for creating a workspace within an existing one, as described earlier.

Once you are done fixing or remediating, you can easily re-deploy an improved version of your model.

Conclusion

Openness and reproducibility is a core Domino offering. With checkpoints added to Durable Workspaces, we expect organizations to foster a collaborative environment where they can reuse past rework and increase user productivity. Also, model-driven businesses can leverage Domino to empower them to proactively monitor and remediate models as needed for their constantly evolving business.

Domino is the Enterprise MLOps platform that seamlessly integrates code-driven model development, deployment, and monitoring to support rapid iteration and optimal model performance so companies can be certain to achieve maximum value from their data science models.