Red Hat OpenShift the most mature and stable Kubernetes solution on the planet
Use Cases and Deployment Scope
We use Red Hat OpenShift as a flexible MLOps platform through OpenDataHub, enabling streamlined model training, tracking, and deployment workflows. It serves as the backbone for our AI Inference Server, allowing us to scale and manage containerized inference endpoints efficiently. Additionally, Red Hat OpenShift hosts our IBM Qiskit development environment via JupyterHub, supporting quantum computing research and prototyping. This setup addresses challenges in deploying reproducible ML pipelines, managing compute resources, and integrating emerging technologies like quantum computing. The scope includes AI/ML development, automated deployment, and hybrid cloud scalability across our research and enterprise infrastructure.
Pros
- Hosting Red Hat OpenShift AT (OpenDataHub)
- LORA Training for Models
- Hositng Inference Systems with MCP Connections
- Running Development Pods for Research Projects
Cons
- The complexity. Some errors occur of systems that cant interact with each other I even dont know run. The system is way to complex in its structure. It is not a OCP issue itself but Kubernetes. To get more adapted, it must be much more integrated and stable.
- The UI is part of the Red Hat OpenShift Container Platform. It should also be on the Red Hat OpenShift Kubernetes Engine (in a simpler way)
- Update Process is failing way too often. There are always issues.
- The User enforcement cant be used in our environment. We need root in pods per standard. This is quite complicated in Red Hat OpenShift.
Return on Investment
- As a research-focused organization, traditional Time-to-Market isn't a key metric for us—but Red Hat OpenShift has significantly expanded our ability to explore and prototype novel AI and quantum simulation workflows without infrastructure bottlenecks.
- The integrated OpenDataHub and JupyterHub environments have improved our productivity by providing a centralized, scalable platform for AI model development and quantum computing experiments.
- Red Hat OpenShift’s strong security model and operator lifecycle management have allowed us to safely experiment with cutting-edge technologies while maintaining a stable and compliant infrastructure.
- While operational costs are higher than simpler setups, the flexibility and innovation it enables have delivered strong research ROI.
Alternatives Considered
HPE Ezmeral Data Fabric (MapR) and HPE Ezmeral Machine Learning Ops
Other Software Used
Proxmox VE, VMware vSphere, Docker, Azure AI Studio

