DH2i adds three Red Hat OpenShift use cases for production AI

Infrastructure as the differentiator for production AI

Red Hat argues that building an AI application is a milestone, but running it reliably against business-critical enterprise data is the harder task. As organizations move retrieval-augmented generation (RAG) and AI assistants from the lab to production, the underlying infrastructure becomes the main differentiator, according to the post.

The post says such models need continuous access to trusted data, the ability to coexist with legacy systems, and a security posture that protects sensitive information across hybrid environments without compromising network integrity.

DH2i provides software for highly available, securely connected SQL Server and application environments across hybrid infrastructure. Its portfolio includes DxOperator, which automates highly available SQL Server Availability Groups on Kubernetes and Red Hat OpenShift, DxEnterprise, for SQL Server high availability across Windows, Linux and containers, and DxOdyssey, for application-level zero trust connectivity across hybrid and multicloud environments.

Red Hat says OpenShift brings AI applications to where the data lives, which matters because of data gravity, AI security and data sovereignty. That lets organizations deploy AI workloads alongside financial and operational data in existing SQL Server databases across hybrid environments.

Always-on RAG for financial teams and a phased path for the public sector

The first use case, Keeping Financial RAG Applications Always On, addresses the risk that a database holding vector embeddings and metadata goes offline. In that case, the post says, the LLM cannot provide grounded answers. The approach uses SQL Server 2025 to store structured data and vector embeddings, while DxOperator automates the lifecycle of highly available SQL Server Availability Groups on OpenShift.

DxOperator can deploy synchronous Availability Group replicas. With synchronous commit, SQL Server waits for transaction log records to be hardened on a synchronized secondary replica before completing the transaction. DxEnterprise adds database-aware monitoring and fully automated failover, which DH2i reports can occur in seconds.

The second use case, Keeping Public Sector AI Services Resilient, targets government agencies that often rely on decades of case records stored in legacy SQL Server environments. Because migrating everything to a cloud-native architecture at once is often impractical, it proposes a "crawl-walk-run" strategy:

• Crawl: rehost existing Windows VMs on Red Hat OpenShift Virtualization.
• Walk: progressively replatform SQL Server workloads from Windows to Red Hat Enterprise Linux.
• Run: fully modernize into containerized SQL Server Availability Groups using DxOperator and DxEnterprise on OpenShift.

DxEnterprise includes the security and networking capabilities of DxOdyssey, so each stage is protected by application-level zero trust connectivity, according to the post. The phased approach is meant to let agencies evolve skills and infrastructure without interrupting data flow to their AI services.

Zero trust access for AI fraud investigation

The third use case, Secure AI Fraud Investigation, deals with financial services data that is often fragmented across private clouds, legacy data centers and customer databases. The post says centralizing it is often impossible, while granting broad network access to AI workloads increases the attack surface.

Unlike traditional VPNs or broad network access, DxOdyssey limits connectivity to specific authorized applications and endpoints, which reduces unnecessary network exposure while preserving access to the needed data, according to the post. Institutions can implement application-level Zero Trust Network Access (ZTNA) by creating micro-tunnels between the AI service on OpenShift and specific authorized data endpoints. Sensitive data stays in place while AI analysts get the information needed to identify suspicious activity.

Red Hat says these micro-tunnels complement OpenShift's native platform security, including zero trust workload identity (SPIFFE/SPIRE ZTWIM), network segmentation (RHACS) and Confidential Containers (CoCo) for hardware-encrypted memory execution.

What it means for organizations

The post's message is that moving AI into production requires looking beyond the model to the platform and data foundation. Red Hat positions OpenShift as the enterprise application platform for production AI at scale, and says that integrating DH2i's products gives organizations added flexibility to maintain SQL Server high availability, streamline modernization and secure hybrid connections.

For teams that already hold critical data in SQL Server, the use cases offer reference approaches for keeping AI services supplied with that data while modernizing at their own pace. The post says this lets teams modernize on their own terms without starting over. The use cases are available in the Red Hat Ecosystem Catalog.