In 2021, Red Hat released Red Hat OpenShift Data Science (RHODS) as a Tech Preview (TP) product, available as a managed service offering on select cloud platforms. Earlier this year Red Hat officially made RHODS Generally Available (GA), which included releasing support for RHODS as a self-managed service offering, bringing the data science tools to even more OpenShift users.
With the growing number of Red Hat customers seeking to expand their data science capabilities, the Data Science & Edge Practice at Red Hat created the RHODS Platform Engagement and the MLOps Foundation Engagement service offerings to help meet our customers wherever they are on their data science journey.
The RHODS Platform Engagement focuses on helping customers get up and running with RHODS and integrate it with their other enterprise tools. The MLOps Foundation Engagement is designed to help customers that have trained a machine learning model and are hoping to leverage MLOps to move beyond a data science experiment, and to create a production ready ML service.
This article will focus on the Red Hat OpenShift Data Science Platform Engagement. To learn more about the MLOps Foundation Engagement, see the companion post here.
The RHODS Platform Engagement focuses on establisning a data science platform on OpenShift in any environment, from cloud based clusters to fully disconnected on-prem environments.
Architecture sessions will help to identify any critical integration points in existing systems, or additional tooling needed to meet each customers unique requirements. The Red Hat team will:
- Work with the customer to determine the tools required for their AI/ML platform.
- Perform an architecture review to understand the client’s environment and determine how RHODS will be integrated.
The Red Hat team will preform the deployment and configuration of all components defined in the architecture design sessions.
- Work with customer to appropriately size configurations, integrate with data systems, and configure hardware accelerators such as GPUs.
- Install, configure, and integrate data processing tools such as Spark and Ray.
- Integrate with other AI/ML tools outside of RHODS.
Customize components of RHODS platform based on architecture design sessions.
- Develop custom pipelines for:
- model training
- data integration
- Create custom notebook images.
Red Hat prioritizes helping to bridge the Cluster Administration/Infrastructure teams and Data Scientists as part of a core foundation of this engagement. Red Hat will work directly with both organizations to help enable your teams for success to maintain the platform moving forward and assist data scientists to get the most of the tools.
After establishing the platform, Red Hat will work directly with Data Scientists to help onboard them to the platform and provide customizations to help improve the user experience.
The RHODS Platform Engagement is available in a number of different sizing options to meet specific customer needs including PoC and dev cluster setup, or larger multi-cluster setups.
A basic Proof of Concopt (PoC) engagement will usually start at 6 weeks of time with Red Hat providing two part time resources (Project Manager and Architect) and one full time resource (Senior Consultant). A PoC engagement will generally target a smaller cluster, possibly a managed cluster from a cloud provider such as ROSA (Red Hat OpenShift Service on AWS) or ARO (Microsoft Azure Red hat OpenShift) environments. The goal of such engagements is generally to evaluate tooling and provide some basic customizations to validate the RHODS platform within your own environment for a small number of users (1-5 users).
A longer term engagement generally start with a minimum 8 weeks with Red Hat providing two part time resources (Project Manager and Architect) and two full time resources (Senior Consultant) and can go up to four months of time. In a longer term engagement, Red Hat is able to provide assistance with a higher level of engagement with existing systems, and may include deployments to multiple clusters, such as a dev, test, and production cluster. Longer term engagements may also involve a larger number of end users, and multiple data science teams.
Customers are required to have a functioning OpenShift cluster prior to the start of the engagement with a dynamic storage provisioning such as OpenShift Data Foundations or the VMWare CSI Driver Operator.
The RHODS Platform Engagement does not require that you have any functioning machine learning models for this engagement, and Red Hat is happy to meet you wherever you and your team are at on your Data Science journey.
For customers looking to move beyond model experimentation and develop strategies for deploying models to production, the MLOps Foundation is a great follow up engagement to the RHODS Platform Engagement.
If you or your business are interested in AI/ML Services offered by Red Hat, speak to your Red Hat account team, or reach out via https://www.redhat.com/en/contact?contact=sales.