ADR-26-02: Oracle RDBMS usage in the REDs
Status |
Accepted |
Impact |
Medium |
Demand Owner |
Alex Weigel |
ADR Contributors |
Alex Weigel, Thomas Thies, Timothy Eyres, Marius Quadflieg |
Reviewers |
Architecture Council |
Informed |
Architecture Council, Heads of CSCoE-DDC |
Publication Date |
April 14th, 2026 |
đ Background
On November 28th 2025, the Data Stores team of RDT announced that Roche is transitioning Oracle Databases into a legacy, maintenance-only phase and standardizing on PostgreSQL for all future developments.
-
Existing Oracle systems (up to v19.29c) will continue to run unchanged, but no new Oracle applications or upgrades are offered on-premise
-
Oracle support will move to a third-party provider, focusing on security and stability only
-
PostgreSQL is now the strategic database platform for all new and modernized solutions
Oracle remains supported short- to mid-term, but teams should avoid new usage and plan for gradual migration to PostgreSQL.
âšī¸ Context
The pRED digital landscape has historically been heavily relying on Oracle database technology for relational data, with over a hundred applications running on Oracle. Amongst these are many foundational systems such as TaPIR (Large Molecule Registration), IRCI (Small Molecule Registration), BDS (Assay Data Warehouse), IDRM (BioProcess Warehouse), or RedPanda (Project-Based Access Management). In addition, pRED heavily relies on Oracle APEX for Low-Cost and Rapid Application Development.
The gRED digital landscape is more diverse and has fewer dependencies on Oracle. Most notable is SMDI, which is the assay data warehouse for small molecules and includes many LIMS components. Also it can be considered as the IRCI counterpart on gRED.
đ Consequences
The Oracle decision made by RDT presents a significant risk to existing solutions in the REDs utilizing Oracle instances hosted on-premise by RDT. Key concerns include:
-
Reduced Support: Oracle support teams have been downsized.
-
Version Stagnation: The Oracle version on-premise is fixed at 19c, limiting access to major enhancements.
-
Compliance Risk: Extended support for Oracle 19c is set to expire by the end of 2029, posing a compliance issue.
-
Cost Implications: RDT may ultimately charge back license costs to business units that continue to use the on-premise Oracle solution.
đ¤ Outcome / Decision
Oracle usage is permissible only under specific conditions related to whether you are implementing a new application or managing an existing solution.
New Solutions to be introduced into the RED landscape (avoid Oracle RDBMS)
New solutions should adhere to the strategic direction of RDT by avoiding Oracle usage and utilizing recommended strategic offerings. This includes no new schemata on existing Oracle onPrem instances.
-
Avoid vendor solutions that have a hard dependency on Oracle.
-
SaaS: Vendor-managed. Ok, provided the vendor holds the Oracle license.
-
IaaS: Oracle to be avoided
-
Paas: Oracle to be avoided
-
-
Utilize the recommended offerings for structured data
-
On prem: RDT Offering via iCare
-
Cloud: Minerva RDS MySQL, PostgreSQL as well as Aurora via self-service
-
Existing Solutions / Platforms (Conditional Usage)
For existing solutions, the decision to remain on Oracle or migrate must be based on a thorough analysis of costs, efforts, and essential feature dependencies.
-
IT Responsibles (such as Product Family Leads and Technical Leads) are required to determine the optimal long-term solution, taking into account the impact on the related data landscape when making a to stay-or-migrate decision.
-
Should a version beyond Oracle 19c be necessary, the corresponding business owner, in collaboration with the systems operations team, is responsible for managing the migration to the new version of Oracle on the AWS RDS and planning for potential license costs.
-
Exception: Teams should only stick with Oracle if the solution has unavoidable dependencies on:
-
Cartridge dependencies without alternatives.
-
High-Performance features for relational databases that competitor solutions cannot meet.
-
Low-Code Platform APEX
-
Oracle APEX
Oracle APEX will be retained, as it is a crucial tool for low-cost and rapid development of data-centric applications. Hosted on Minerva (AWS), the CS-CoE Oracle APEX Platform offers a strong Return on Investment (ROI) due to low licensing costs (Oracle Standard Edition), high application density per database instance, and minimal operational overhead. Licensing for the database is included in higher compute prices for RDS Oracle. The whole Oracle RDS costs are in the AWS Cloud budget, which is covered by RDT. The business unit only sees the costs but doesn’t have to pay them.
APEX and AI: AI tools accelerate application development but primarily address code generation, whereas Oracle APEX provides a complete, low-operational runtime platform. Oracle is actively embedding AI capabilities into APEX (AI-assisted development and declarative integration of generative AI services).
As a result, AI-based solutions built outside such integrated platforms typically require additional effort in operations, maintenance, and compliance. Therefore, APEX remains a preferred option for low-cost, data-centric enterprise applications, with AI serving as a complementary productivity enhancer. This position should be re-evaluated within the next 1â2 years as AI capabilities mature.
Recommendations
All applications currently using Oracle RDBMS should plan for migration to be finished by the end of 2029, unless they qualify for an exception. This deadline is critical because Oracle 19c will reach its end-of-life, leading to increased compliance risks due to the absence of security patches and official support. Therefore, this necessary migration must be integrated into the life-cycle planning for every affected application.
đ Resources
-
First holistic analysis and estimates conduct in legacy D&A organisation in 2024.