On the heels of the general availability of the Red Hat OpenShift Data Science cloud service a little more than a month ago, we’re excited to announce the GA of a self-managed software version. 

Red Hat OpenShift Data Science (RHODS) is an easy-to-configure MLOps platform for building and deploying AI/ML models and intelligent applications. OpenShift Data Science delivers the core environment where data scientists can train models using notebook images, which include frameworks such as TensorFlow and PyTorch among others with access to both software-defined CPU accelerators and GPUs.  This new version of OpenShift Data Science also includes the enhanced model serving capabilities introduced in December. Customers can extend the RHODS environment with their own notebook images and optional partner technologies and add other open source AI/ML technologies.

Red Hat OpenShift Data Science provides pre-integration of several optional technology partner offerings including Anaconda Enterprise, IBM Watson Studio, Intel OpenVINO and AI Analytics Toolkit, Pachyderm, and Starburst Galaxy (for the cloud service). In addition, any of the more than 30 AI/ML software partners that have certified OpenShift operators can also be used in combination with RHODS.

While the cloud service provides more of a hands-off IT ops experience, especially for customers who want to leave the application platform infrastructure and tooling updates to Red Hat SRE experts in concert with our cloud partners like AWS, a self-managed version appeals to those who:

  • Want to keep the entire data prep, model development, and model deployment closer to the data - on-prem or even at the enterprise edge
  • Are restricted from deploying in the cloud by compliance requirements
  • Are users of Red Hat’s premier application platform, Red Hat OpenShift, but have not yet adopted Red Hat cloud services (OpenShift Dedicated or Red Hat OpenShift Service on AWS) 

Our approach is to keep the functionality the same for the cloud service and self-managed product.  We also plan to iterate the self-managed version rapidly to keep the release cycles in sync with the cloud service.  Of course, there will be some slight differences.  For example, air-gapped support for disconnected environments will be introduced in the self-managed version in late January and will appeal to government customers and others requiring limited connectivity.

Most of all, we feel offering both a managed cloud service and traditional software AI platform is consistent with Red Hat’s core focus on giving customers choice.  Develop models in the cloud and deploy on prem.  Develop models on prem and deploy models at the edge or cloud.  Your choice. In addition to providing a powerful application platform powered by Kubernetes, the ability for OpenShift (and now OpenShift Data Science) to run on multiple footprints provides a hybrid MLOps environment that collaboratively brings IT, data science and application development teams together.

RHODS relieves customers of the burden of wiring these AI/ML technologies together and managing separate life cycles on their own. It simplifies the management of the platform for IT Operations, enabling platform engineers to create configurations for their data scientists and application developers that can scale up or down and be administered with less effort. 

Early customers have told us they really like how easy it is for IT Ops to manage the environment for their data science and application development teams. They have commented that the admin UI capabilities make it easier to tailor the solution to meet their needs. Early customers also appreciated the ability to get their users up and running quickly with familiar data science tools and their own custom notebook images. 

Don’t take our word for it.  Give it a spin.  We have a developer sandbox, complete with sample learning paths, for data practitioners willing to get their feet wet.