14 · Choosing your Oracle Database@Azure service
The provisioning lessons in this module walk through Oracle Exadata Database Service on Dedicated Infrastructure — creating the infrastructure, a VM cluster, and a database. That is one of several database services you can run under Oracle Database@Azure. This page maps the whole family so you pick the right service for a workload before you start provisioning.
Core message
Oracle Database@Azure is a family of database services, not a single product. Under one onboarding, one Azure bill, and one delegated subnet, you can deploy Oracle databases on Oracle-managed hardware inside Azure datacenters in several shapes. They differ on three axes: the infrastructure model (dedicated hardware reserved for you, or shared elastic capacity), who operates the database (you tune and patch it, or Oracle runs it as a self-driving Autonomous service), and cost and entry point. Choose by how much you want to manage and how predictable the workload is. Microsoft and Oracle have also rebranded the service to Oracle AI Database@Azure, reflecting the AI-capable 23ai database release — the same service and deployment shapes, a newer name and database version.
The service family
All of these run on Oracle-managed hardware in Azure datacenters, reached through a delegated subnet, and all support Oracle 19c and 23ai.
| Service | Infrastructure | Who runs the database | Typical fit |
|---|---|---|---|
| Exadata Database Service on Dedicated Infrastructure | Dedicated Exadata rack reserved for you | You (Oracle runs the hardware) | Mission-critical, latency-sensitive, regulated workloads needing dedicated isolation and full control |
| Exadata Database Service on Exascale Infrastructure | Shared, elastic Exadata | You (Oracle runs the hardware) | Exadata performance at a lower entry point; right-size per department, market, or budget |
| Base Database Service | Shared general-purpose VMs with network-attached storage | You | Single-instance Oracle with your choice of database edition, at lower cost |
| Autonomous AI Database on Dedicated Exadata | Dedicated Exadata | Oracle (self-driving) | Autonomous automation with the isolation of dedicated infrastructure |
| Autonomous AI Database (Serverless) | Shared Exadata | Oracle (self-driving) | Variable or cloud-native workloads that scale up and down with minimal administration |
Two adjacent services round out the family: Oracle GoldenGate for replication and migration, and Oracle Database Autonomous Recovery Service for managed backup with stricter recovery objectives than plain object-storage backup.
Dedicated versus shared infrastructure
- Dedicated (Exadata Dedicated, Autonomous on Dedicated) — your own Exadata infrastructure with complete isolation. You pay an upfront subscription for the infrastructure plus pay-per-use for the CPUs allocated to the VM cluster. Performance is predictable because nothing else shares the rack.
- Shared (Exascale, Base, Autonomous Serverless) — elastic capacity you scale by ECPU and storage, with a lower entry cost. Exascale and Autonomous still deliver Exadata-class performance on the shared platform.
You run it, or Oracle runs it
- Exadata Database Service and Base Database Service — cloud automation creates the database and applies infrastructure maintenance, but you tune the database, apply Database, Grid Infrastructure, and Guest OS patches on your own schedule, and configure any extra monitoring.
- Autonomous AI Database — self-driving. Oracle automates provisioning, scaling, tuning, patching (with no downtime), and backups. You focus on the data and the application rather than database administration.
A quick way to choose
- Need dedicated isolation and full control for a mission-critical Exadata workload → Exadata Database Service on Dedicated Infrastructure.
- Want Exadata performance with a lower entry point or elastic sizing → Exadata Database Service on Exascale Infrastructure.
- Running single-instance Oracle and want the lowest cost with a choice of editions → Base Database Service.
- Want the least operational effort with elastic, pay-per-use scaling → Autonomous AI Database (Serverless).
- Want Autonomous automation on isolated, dedicated hardware → Autonomous AI Database on Dedicated Exadata.
If you need Oracle RAC, that points to the Exadata services — Base Database Service is single-instance.
Where the services are available
Each service is offered per Azure region paired to an OCI region. Most services are available across the published regions, but a few vary by region. Always confirm the current list on the Microsoft Learn Region availability for Oracle AI Database@Azure page before you commit, because the list changes as Oracle and Microsoft add regions.
The naming: Oracle AI Database@Azure
You will see two names for the same thing. This module uses Oracle Database@Azure (OD@A), the name most customers still say. Microsoft and Oracle now brand the service Oracle AI Database@Azure, reflecting the AI-capable 23ai database release and in-database AI features such as AI Vector Search. The service and the deployment shapes are unchanged — only the name and the database version have moved forward. Treat the two names as the same service, and do not assume the name change alone implies a new capability in a given workload.
Customer value
- One onboarding, many shapes — a single purchase, Azure bill, and delegated subnet cover every service, so a customer can mix Exadata, Base, and Autonomous without a second commercial motion.
- Right-sized cost — dedicated infrastructure for predictable mission-critical workloads, shared and serverless options for variable ones, so spend follows the workload.
- A managed option for every level of control — from full DBA control on Exadata Dedicated to a fully self-driving Autonomous database.
- Consistent governance — every shape stays in the customer's Azure VNet, on the Azure bill (counting toward MACC), and under Microsoft Entra sign-in.
Risks and considerations
- Availability varies by region — confirm the specific service is offered in the target region before you design around it.
- Autonomous takes logical migration only — you cannot move into an Autonomous database with a physical RMAN or Data Guard method; use Data Pump, a database link, or GoldenGate.
- Dedicated infrastructure carries an upfront subscription — size it to real demand rather than a peak you rarely hit.
- Base Database Service is single-instance — it has no RAC; use the Exadata services when you need RAC.
- The "AI Database" naming is still rolling out — plan for both names appearing in portals and documents during the transition.
Terms to remember
- Exadata Database Service on Dedicated Infrastructure — Oracle-managed Exadata rack reserved for you; you operate the database.
- Exadata Database Service on Exascale Infrastructure — Exadata performance on shared, elastic infrastructure with a lower entry point.
- Base Database Service — single-instance Oracle on shared general-purpose infrastructure with a choice of editions.
- Autonomous AI Database — Oracle's self-driving database, available serverless (shared) or on dedicated Exadata.
- Oracle AI Database@Azure — the current name for the Oracle Database@Azure service family, aligned with the 23ai release.
"When a customer says 'we want Oracle on Azure,' I do not jump straight to Exadata. I frame it as a menu. If the workload is mission-critical and regulated and they want their own isolated rack with full DBA control, that is Exadata Database Service on Dedicated Infrastructure. If they want that Exadata performance but a lower entry point and elastic sizing, that is Exascale. If it is a single-instance database and cost is the driver, Base Database Service fits. And if they would rather not run the database at all — no tuning, no patching, scale on demand — that is Autonomous, either serverless for variable workloads or on dedicated hardware if they need the isolation. Same onboarding, same Azure bill, same delegated subnet for all of them — so the decision is really just how much do you want to manage, and how predictable is the workload. One name note I always give: they will see Oracle AI Database@Azure in the portal now — that is the same service, renamed around the 23ai release."