When a core database goes down, the cost is measured in lost revenue and regulatory exposure — not in IT tickets. For Slovak and Czech organisations under NBS, ČNB or NÚKIB oversight, business continuity isn’t a nice-to-have; it’s a requirement. Oracle’s Maximum Availability Architecture (MAA) is the validated blueprint for getting there, and OCI gives you the building blocks. Here’s how it fits together — and how to pick the right level without overspending.
The MAA tiers: from Bronze to Diamond
MAA was, for years, a four-tier model. It now has five — Oracle added a Diamond tier for extreme-availability and real-time AI workloads. Each tier maps a service-level target (RTO/RPO) to a concrete set of technologies, so you buy exactly the resilience you need.
RTO/RPO are Oracle MAA design targets, not guarantees.
A quick orientation:
- Bronze — single instance plus RMAN backups. Fine for dev/test; recovery in minutes to hours.
- Silver — local HA via 2-node RAC or a local Data Guard standby, with Application Continuity masking node failures (RTO under ~60s).
- Gold — adds regional DR with Active Data Guard and Fast-Start Failover; the baseline for business-critical systems (DR RTO under ~5 min, zero/near-zero data loss).
- Platinum — Exadata plus GoldenGate, for mission-critical, near-zero-downtime operations.
- Diamond — the new apex: Oracle AI Database 26ai on Exadata with RAC, Active Data Guard and GoldenGate, targeting failover in seconds for the most demanding (often AI-driven) workloads.
How OCI’s primitives map to resilience
You can’t design DR on OCI without knowing three building blocks:
- Fault Domain — hardware isolation within one data center (independent racks, power, switches). A 2-node RAC cluster is automatically split across Fault Domains, so a rack failure or a hypervisor patch only touches one node.
- Availability Domain (AD) — a fully isolated data center within a region (own power, cooling, network). Spreading across ADs survives a facility-level incident.
- Region — geographically distant (often hundreds of km). This is where cross-region disaster recovery lives.
OCI also triple-mirrors the network-attached block storage behind RAC, so the surviving node keeps full access to the data files even if a storage controller fails — without the manual storage gymnastics on-prem clustering demands.
RAC, Data Guard, Active Data Guard: what protects what
These are different tools for different failures:
- RAC is local high availability — active-active instances on multiple nodes. It survives node/hardware failure and enables rolling, zero-downtime patching. On OCI it runs on flexible VM shapes (minimum 2 OCPUs per node); note that bare-metal shapes can’t be clustered.
- Data Guard / Active Data Guard is replication — a block-for-block standby in another AD or region. Sync transport (Maximum Availability) → RPO 0; async (Maximum Performance) → RPO of seconds, used for cross-region DR to avoid WAN latency. Active Data Guard keeps the standby open read-only, so it earns its keep running reports and (on 26ai) even in-database AI inference.
- Switchover vs failover. A switchover is planned and lossless; a failover is unplanned and, with Fast-Start Failover plus an observer, automatic — cutting RTO to seconds. Far Sync lets you keep RPO 0 over long distances by buffering synchronous redo at an intermediate hop.
This is the design work behind our RAC & Data Guard service — the difference between “we have a standby” and “we have a failover we’ve actually tested.”
Managed vs DIY: the honest trade-off
Autonomous Database delivers MAA Gold automation out of the box: Autonomous Data Guard gives a local standby with RTO ~2 minutes and RPO 0, fully automated, and the standby is billed only at baseline CPU (your auto-scaling peaks on the primary aren’t double-charged). There’s also a cheaper Backup-Based DR (no live standby) with a higher RTO of roughly “1 hour + 1 hour per 5 TB.”
DIY on Base Database or Exadata gives you total control — and total responsibility: peering VCNs, configuring the Data Guard Broker, running the FSFO observer, scripting application failover, scheduling rolling patches. More power, more ways to get it wrong under pressure. The right choice depends on how much control your applications and compliance regime actually require.
Cross-region DR and the EU Sovereign Cloud
For SK/CZ enterprises, the strongest pattern keeps data inside the EU’s legal boundary. The OCI EU Sovereign Cloud runs in separate realms operated solely by EU-resident staff under EU entities, with two regions — Frankfurt and Madrid. A cross-region Active Data Guard association between them (async transport over private VCN peering) gives you geographic DR where sensitive data never leaves EU jurisdiction — directly addressing GDPR, NIS2 and DORA concerns.
One practical detail that sinks projects: the two regions’ database subnets must use non-overlapping CIDR ranges, or role transitions break on IP conflicts.
The cost and licensing edge
Resilient designs replicate a lot of redo data between regions — which is exactly where OCI’s economics help:
- Egress: OCI includes 10 TB/month of outbound transfer free and charges far less beyond it, so cross-region redo shipping doesn’t carry the punitive transfer bills typical elsewhere.
- Licensing: under Oracle’s cloud policy, one EE Processor licence covers two OCPUs (≈4 vCPUs) on OCI, versus only two vCPUs on AWS or Azure. A workload needing 32 vCPUs needs 8 EE licences on OCI but 16 on AWS/Azure — and a DR standby doubles whatever that number is. For HA/DR specifically, that halving compounds.
The headline figures change over time, so confirm current pricing — but the structural advantage for Oracle-licensed HA/DR is real, and it’s why a proper TCO comparison usually favours OCI for these workloads.
How to choose a tier
Match the tier to the business cost of downtime, not to ambition:
- Non-critical / dev → Bronze.
- Standard production that must survive a node failure → Silver.
- Business-critical needing regional DR and ~zero data loss → Gold.
- Mission-critical, near-zero downtime even during maintenance → Platinum.
- Extreme availability / real-time AI → Diamond.
Bottom line
OCI gives you every MAA building block — RAC, Data Guard, Active Data Guard, sovereign-region DR — at licensing and egress economics that favour Oracle workloads. But resilience is a design, not a checkbox: the tier, the transport mode, the failover automation and the testing all have to match your real RTO/RPO and compliance targets.
That design — and a failover you’ve rehearsed — is what we build, independently and without reselling infrastructure. If continuity is on your roadmap, let’s talk.
