1z0-1147 Oracle AI Database@Azure Architect Professional Questions and Answers

Oracle Database Autonomous Recovery Service supports Long-Term Retention (LTR) backups for organizations that need to preserve full database backups for compliance, regulatory, legal, or business requirements. The correct destination is the OCI Object Storage Infrequent Access tier .

The question source specifies that the LTR backup initially remains in the standard storage tier and transitions after the first 24 hours to Infrequent Access storage for the remainder of its retention period. Oracle's Recovery Service documentation independently confirms that long-term retention backups are stored in the Object Storage Infrequent Access tier and can be retained for periods extending up to ten years.

The Infrequent Access tier is appropriate because LTR backups are retained for extended periods but are generally restored less frequently than operational backups. Archive Storage is not the managed destination used by Recovery Service for this workflow. Azure Archive Blob Storage is likewise not the Oracle-managed OCI storage tier identified for these backups, and there is no "OCI Legacy Backup" storage tier.

Therefore, D is correct.

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Autonomous Recovery Service, LTR backups, retention and storage lifecycle.

===============

ZDM manages migrations as persistent scheduled jobs and assigns each job a unique job ID. Administrators must manipulate these jobs through the supported ZDMCLI lifecycle commands rather than modifying ZDM repository or log files manually.

Oracle documents the command zdmcli abort job -jobid < job_id > for terminating a specified ZDM job. When an existing job context is preventing a new migration workflow from being established for the same source, the supported administrative approach among the available options is to identify the existing job and abort it through ZDMCLI.

Restarting the ZDM service does not remove the persistent job record; the job metadata is maintained by ZDM rather than existing only in the running daemon's memory. Manually deleting files from ZDM directories is unsupported and risks corrupting or desynchronizing job state. Likewise, the proposed -ignore ALL option is not a supported mechanism for bypassing ZDM's migration-job lifecycle controls.

Consequently, B is the correct operational command because it acts on the existing migration through the supported ZDM control interface.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — ZDMCLI job management, query/abort operations and migration workflow lifecycle.

===============

Elastic resource allocation is a fundamental design characteristic of Oracle Exadata Database Service on Exascale Infrastructure . A VM requires a minimum of 8 enabled ECPUs at initial provisioning, but its compute allocation can subsequently be increased in 4-ECPU increments .

Oracle's Exascale architecture draws compute resources from Oracle-managed shared infrastructure rather than requiring the customer to purchase or provision dedicated database servers. This permits a customer to start with a relatively small footprint and increase capacity as workload requirements develop.

The workload therefore does not need to be migrated to Dedicated Infrastructure merely because additional processing power is required. Nor is it necessary to destroy and rebuild the VM Cluster or create an entirely new Exadata Infrastructure resource. Exascale is specifically intended to avoid those fixed-capacity operational patterns.

Oracle additionally supports online scaling of Exascale storage independently through the associated Database Storage Vault, reinforcing the service's elastic consumption model.

Therefore, D is correct.

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Exascale elastic ECPU scaling.

===============

Oracle Autonomous AI Database is specifically designed for workloads where elasticity and minimal administrative overhead are primary requirements. Oracle describes it as a fully autonomous, self-governing database service that handles provisioning, backups, patching, upgrading, performance management, and resource scaling without traditional database-administration effort.

For unpredictable workloads, compute auto scaling is particularly significant. When enabled, Autonomous AI Database can automatically consume additional CPU and I/O resources as workload demand rises without requiring an administrator to manually resize the database. Storage can likewise be scaled and can use separate auto-scaling functionality.

Exadata Database Service on Dedicated Infrastructure provides extremely high performance and consolidation capabilities but retains customer database-management responsibilities. Exascale Infrastructure improves Exadata's elastic infrastructure model but is still not equivalent to Autonomous Database's fully managed, self-governing operational model. Base Database Service similarly requires substantially more customer administration.

Therefore, a cloud-native application with volatile usage and a requirement for essentially zero database administration maps directly to Oracle Autonomous AI Database .

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Autonomous AI Database, elastic scaling, automated management and serverless operations.

===============

The standard Oracle Autonomous AI Database Serverless resource-allocation interface permits the administrator to configure the base compute allocation up to 512 ECPUs using the ECPU slider. This value represents the configured base ECPU count for a single database instance rather than the potential temporary compute consumption obtainable through compute auto scaling.

Oracle distinguishes configured ECPUs from auto-scaled usage. For eligible Autonomous AI Database configurations, compute auto scaling can allow the database to consume additional processing capacity above the configured base according to workload demand. For example, a database configured with 512 ECPUs may be capable of consuming substantially more effective processing capacity when auto scaling is enabled, subject to the service and licensing model.

The 1024-ECPU option therefore does not represent the normal single-instance slider ceiling addressed by this question. The 256- and 128-ECPU values are valid intermediate configurations but are below the maximum.

Consequently, A. 512 ECPUs is correct.

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Autonomous AI Database Serverless provisioning, ECPU configuration and compute scaling.

===============

Oracle AI Database@Azure has a default regional service limit of five delegated subnets . Once a deployment requires more than five delegated subnet links in a region, the customer must request a service-limit increase.

This aggregate regional limit must not be confused with the separate VNet design rule. An individual Azure VNet can contain a maximum of one delegated subnet for Oracle AI Database@Azure, but an organization can have several VNets and therefore several delegated subnets across the same region up to the default service limit.

Oracle identifies the relevant service-limit metadata as:

Service: Multicloud

Resource: Delegated Subnet Multicloud Links

Limit name: azure-delegated-subnet-count

The limit operates at regional scope. Therefore, creating additional peerings or Route Servers does not bypass it; the correct approach is to request additional service capacity through Oracle's service-limit process.

Values of 1, 3, and 10 do not represent the default regional limit.

Therefore, A. 5 is correct.

Study Guide reference: Configure Oracle AI Database@Azure Networking — delegated subnet quotas and service-limit increases.

===============

Azure User-Defined Routes provide explicit Layer-3 forwarding control for traffic entering or leaving an Oracle AI Database@Azure delegated subnet. Oracle's documented network topologies use UDRs to direct application, on-premises, and database traffic through prescribed next hops such as an Azure Firewall or Network Virtual Appliance. This enables architects to construct custom network routing paths between on-premises systems and Exadata databases and exercise granular routing control across hybrid infrastructure .

In a Hub-and-Spoke deployment, for example, UDRs on application spokes can force database-bound traffic through the hub firewall, while corresponding routes on the Oracle AI Database@Azure spoke provide a symmetric return path. This is an architectural routing control, not a dynamic-routing protocol. ExpressRoute path selection is governed by BGP and ExpressRoute configuration rather than by a UDR automatically choosing between circuits.

Likewise, UDRs do not override Oracle-managed control-plane connectivity or scheduled lifecycle-management routes. Those functions remain under the service architecture.

Therefore, A and C are the supported routing capabilities. Study Guide reference: Configure Oracle AI Database@Azure Networking — UDRs, delegated subnets, hybrid connectivity, Hub-and-Spoke routing.

===============

For ZDM physical online migrations using direct data transfer , Oracle's Maximum Availability Architecture pattern uses RMAN's Restore From Service capability. Rather than first creating backup pieces in a staging location, transporting them, and then restoring them on the target, RMAN restores the target database directly across the network from the source service.

This design reduces intermediate storage requirements and simplifies the migration pipeline. Once the target database has been instantiated, ZDM automatically configures Oracle Data Guard and uses redo transport to maintain synchronization until the planned switchover. The combination of Restore From Service and Data Guard is the central technical pattern Oracle documents for physical online migration to Exadata Database Service on Oracle AI Database@Azure.

Although DUPLICATE ... FROM ACTIVE DATABASE is a legitimate RMAN cloning capability in other contexts, it is not the named method documented for this ZDM direct-data-transfer workflow. Compressed backup-set streaming and incremental image-copy pushing also imply different backup/staging mechanisms.

Therefore, Restore From Service is the MAA-aligned RMAN method represented by the scenario.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — MAA, ZDM Physical Online migration, RMAN direct data transfer and Restore From Service.

===============

When Oracle Autonomous AI Database Serverless is provisioned through Oracle AI Database@Azure, the administrative database account follows the standard Autonomous Database account model. The Azure creation workflow exposes the Username field as read-only and set to ADMIN .

The administrator therefore supplies the password for the predefined ADMIN account but does not choose or rename the account during database creation. Oracle currently specifies password requirements separately, including minimum and maximum length and character-composition requirements.

The account is not derived from the Microsoft Entra ID identity that deploys the Azure resource. Azure RBAC and Microsoft Entra authentication govern access to the Azure resource-management plane, whereas the ADMIN credential is a database-level account. Similarly, SYS is not presented as the configurable provisioning username for Autonomous AI Database Serverless, and the username is not generated from subscription metadata.

This separation between Azure control-plane identity and Oracle database credentials is architecturally important when designing authentication and administrative access.

Therefore, the correct answer is A .

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Autonomous AI Database Serverless creation, configuration and administrator credentials.

===============

The -eval option invokes the evaluation workflow of Oracle Zero Downtime Migration rather than executing the production migration. Oracle specifically directs administrators to use ZDMCLI migrate database ... -eval to evaluate the migration process, including prerequisite checks, source/target connectivity verification, configuration validation, and applicable pre-migration assessment phases.

The resulting job is an evaluation job rather than a migration job that performs database movement. This distinction is reflected throughout ZDM command handling: ZDM can query evaluation jobs separately from actual migration jobs, and the evaluation workflow is intended to identify prerequisite or configuration failures before production data movement begins.

The option does not create a mock standby or execute a simulated Data Guard switchover. It also does not begin an offline migration and stop before the copy phase. CPAT may participate in migration-readiness assessment depending on ZDM version and command options, but -eval is not an installation command for CPAT.

Consequently, A correctly represents the effect: an EVAL-type workflow performs migration prechecks and validation without executing the actual migration.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — ZDM CLI, evaluation mode, migration validation and prechecks.

===============

Oracle Zero Downtime Migration organizes its Oracle database migration methods into two fundamental workflow families: Physical Migration and Logical Migration . Each family can then have online and offline variants depending on compatibility requirements and acceptable downtime.

Physical migrations operate on the physical database representation and rely on technologies such as RMAN and Oracle Data Guard. Physical Online migration can instantiate a target using Restore From Service and maintain synchronization using Data Guard, while Physical Offline uses RMAN backup and restore.

Logical migrations operate through logical database objects and data. Oracle Data Pump provides the initial export/import mechanism, while Logical Online migration integrates Oracle GoldenGate to capture and replicate ongoing transactions before cutover. Logical workflows are also suitable for many cross-version or cross-platform scenarios that physical workflows cannot support.

"Container-based migration" and "snapshot-based migration" are not ZDM's top-level migration workflow categories.

Oracle's current ZDM documentation explicitly states that ZDM supports both Physical and Logical Migration workflows.

Therefore, B and D are correct.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — ZDM workflow categories.

===============

Oracle's preferred ZDM Physical Online workflow for Oracle Database@Azure supports direct data transfer , specifically to avoid staging large RMAN backups in intermediate object or file storage. In this workflow, ZDM creates the target database using the Restore From Service method.

RMAN streams the required database blocks directly from the source database service to the target environment over the configured network connection. After the target database has been instantiated, Oracle Data Guard maintains synchronization between source and target by transporting and applying redo. ZDM then orchestrates the cutover when the migration reaches the appropriate stage.

This approach is particularly significant for a large database such as the 10 TB system in the scenario because intermediate backup staging would require substantial temporary storage, additional I/O, and backup-transfer operations. Oracle explicitly states that the physical online direct-data-transfer workflow creates the target through Restore From Service and avoids backing up the source to an intermediate storage location.

Backup and Restore therefore describes the physical offline/staged model. Transportable Tablespaces is not the standard ZDM physical-online instantiation mechanism.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — ZDM Physical Online, direct data transfer, RMAN Restore From Service and Data Guard.

===============

The requirement is specifically to place backup recovery points outside the cloud that hosts the production database . Oracle identifies this as the Cross-Cloud Backups multicloud use case.

In a cross-cloud backup architecture, production data hosted in one cloud is backed up to a different cloud provider. Oracle describes this pattern as providing protection against failures, operational mistakes, attacks, or other incidents affecting the primary cloud. Because the secondary copy is located outside the original cloud environment, it also provides an additional isolation boundary that can satisfy organizational or regulatory requirements for off-site backups.

An application/database split-stack design solves a different problem: it places application and database tiers in different clouds to exploit provider-specific capabilities. Horizontal workload distribution spreads active workloads across providers, while production/development separation assigns lifecycle environments to different clouds. Neither inherently guarantees an independently located backup recovery point.

The scenario is therefore a direct match for Oracle's Cross-Cloud Backups pattern.

Study Guide reference: Introduction to Multicloud — Cross-Cloud Backups, resilience, data protection and provider diversification.

===============

The default encryption-key model for an Exadata database on Oracle AI Database@Azure is Oracle-Managed Keys (OMK) . Under this model, Oracle automatically generates the Transparent Data Encryption master encryption key and stores it in an Oracle Wallet secured within the database environment.

Oracle manages the complete lifecycle of this default key, including generation, secure storage, rotation, backup, and restoration. No additional customer-side key-management infrastructure is required.

OCI Vault, Oracle Key Vault, and Azure Key Vault represent customer-managed key alternatives. When an organization selects one of these mechanisms, the TDE master key is managed outside the local Oracle Wallet, enabling centralized customer control, auditing, rotation policy, and separation of duties. However, none is the default unless the customer explicitly configures a customer-managed key architecture.

Oracle's current security documentation explicitly identifies Oracle Wallet as the default encryption-key setting for Exadata Database Service on Oracle AI Database@Azure.

Therefore, C. Oracle Wallet is correct.

Study Guide reference: Secure Oracle AI Database@Azure — TDE, Oracle-managed keys and Oracle Wallet.

===============

Oracle Exadata Database Service on Exascale Infrastructure separates database storage into Exascale Database Storage Vaults . Oracle documents the minimum billed Database Storage Vault capacity as 300 GB per VM Cluster .

The minimum is composed of storage required for system use and the first database. Oracle's current Exascale capacity table allocates 50 GB for system-related usage, including images stored in ACFS, and 250 GB for database use when provisioning the first database. Together, these produce the 300 GB minimum vault capacity.

This value must not be confused with VM file-system storage. Current Exascale guidance specifies a separate minimum VM file-system capacity that varies by database version—for example, 220 GB for a 26ai VM configuration. That explains why 220 GB appears plausible but is incorrect for the Database Storage Vault asked about here.

The vault can subsequently be scaled substantially beyond the minimum; Oracle documents VM Cluster Exascale Vault storage scaling up to 100 TB.

Therefore, the correct answer is D. 300 GB .

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Exadata Database Service on Exascale Infrastructure, Database Storage Vault sizing and capacity.

===============

The two managed backup destinations represented in the question are OCI Object Storage and Oracle Database Autonomous Recovery Service .

OCI Object Storage is the established managed backup destination for Oracle Database services and stores Oracle-managed database backup objects in OCI. Autonomous Recovery Service is Oracle's recommended advanced database-protection platform. It provides database-aware incremental-forever backup processing, centralized protection policies, recoverability validation, retention locking, and optional real-time transaction protection.

For Oracle Database@Azure, Recovery Service can store protected-database backups either in the paired OCI region or, where supported and selected by policy, within the same Azure cloud provider location. Oracle architecture documentation therefore increasingly positions Autonomous Recovery Service as the preferred automatic-backup destination.

The physical Zero Data Loss Recovery Appliance is an on-premises engineered system rather than the managed cloud backup destination in this service. Direct Azure Blob Storage API access is likewise not presented as the standard built-in automatic Exadata backup destination; Azure-hosted Recovery Service storage is accessed through the managed Recovery Service architecture rather than direct customer Blob APIs.

Therefore, C and D are correct.

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Exadata backup destinations and Autonomous Recovery Service.

===============

Oracle's cross-region Oracle Database@Azure reference architecture requires an additional Hub VCN because each OCI VCN can be attached to only one Dynamic Routing Gateway (DRG) at a time.

The Oracle-managed database VCN associated with an Exadata VM Cluster already uses a DRG as part of the Oracle Database@Azure network architecture. Cross-region VCN peering also requires DRG-based remote peering. Because a second DRG cannot simply be attached to the existing database VCN, Oracle introduces a separate Hub VCN with its own DRG. The database VCN is locally connected to the Hub VCN, and the Hub DRGs establish the cross-region remote peering connection.

OCI documentation independently confirms that one VCN can attach to only one DRG, whereas a single DRG can support many VCN attachments. A VCN can also have multiple Local Peering Gateways, so option B is technically incorrect.

The limitation is therefore architectural rather than a restriction caused by Azure subnet delegation or tenancy boundaries.

Study Guide reference: HA and DR — OCI transit networking, Hub VCN, DRG, LPG and cross-region Data Guard.

===============

Logical Online Migration combines an initial logical data load with Oracle GoldenGate change data capture to minimize production downtime.

The initial target database is populated using Oracle Data Pump export and import. Because that data movement can take a substantial amount of time for large databases, Oracle GoldenGate continuously captures transactions committed on the source while the initial load is being performed. Those transactions are subsequently applied to the target, allowing the destination to converge with the production source while the application remains active.

When synchronization reaches the required state, ZDM orchestrates the final migration phases and application cutover. The downtime requirement is therefore concentrated around the final transition rather than the entire export/import window.

Logical Offline uses Data Pump without continuous GoldenGate synchronization. Physical Online uses RMAN-based target instantiation and Oracle Data Guard redo synchronization instead of GoldenGate. Physical Offline similarly relies on physical backup/restore processes and does not use logical change-data capture.

Thus, B. Logical Online Migration is correct.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — ZDM Logical Online, Data Pump and GoldenGate CDC.

===============

For Oracle AI Database@Azure's documented Azure-to-OCI federation architecture, Microsoft Entra ID functions as the enterprise identity provider. Oracle's reference architecture explicitly states that Microsoft Entra ID provides federated identity and access management for Oracle AI Database@Azure.

The OCI IAM Identity Domain remains an essential component because it is the Oracle-side service provider and identity domain that trusts the Microsoft Entra identity provider. However, when the question asks which platform provides the federated corporate identity in the Azure-integrated architecture, the answer is Microsoft Entra ID.

Oracle's onboarding instructions configure an Oracle Cloud Infrastructure Console enterprise application in Entra ID and then configure the OCI identity domain with Microsoft Entra ID as its SAML identity provider. Administrators can subsequently use their corporate Entra credentials for SSO into OCI when Oracle-native management tasks require it.

AD FS represents an older on-premises Microsoft federation service, while Okta is a third-party identity platform rather than the natively documented Azure integration.

Therefore, A is correct.

Study Guide reference: Secure Oracle AI Database@Azure — Entra ID federation, OCI IAM Identity Domains and SSO.

===============

For a cross-region Autonomous Data Guard standby , Oracle specifies a maximum possible data-loss objective of approximately 1 minute . Accordingly, the intended Recovery Point Objective is one minute.

Cross-region protection differs from a local Autonomous Data Guard standby because geographical distance makes fully synchronous zero-data-loss transport impractical for normal cross-region operation. Oracle therefore uses an asynchronous protection model that provides a bounded RPO while maintaining acceptable database performance.

A useful distinction is between RPO and RTO. RPO measures potential data loss, whereas RTO measures how long restoration of service is targeted to take. Oracle's current Autonomous AI Database Serverless SLO documentation continues to specify a 1-minute cross-region RPO . Current documentation has improved the cross-region RTO target to less than 10 minutes ; older material cited approximately 15 minutes. That RTO evolution does not affect the answer to this question because only RPO is being tested.

Zero data loss therefore applies to different protection scenarios, while five and fifteen minutes do not represent the documented cross-region RPO.

Study Guide reference: HA and DR — Autonomous Data Guard local and cross-region recovery objectives.

===============

Oracle AI Database@Azure does not eliminate the need for an OCI tenancy. Oracle explicitly requires an Oracle Cloud account/OCI tenancy during onboarding. A Public Offer creates a new tenancy, while a Private Offer can link an eligible existing tenancy or create one.

The other choices describe genuine architectural advantages. Oracle deploys its database infrastructure inside isolated areas of selected Microsoft Azure data centers, which places Oracle database resources physically close to Azure applications. This enables low-latency application-to-database communication through Azure networking rather than forcing production data traffic across the public internet to a distant OCI region.

Oracle also configures a managed connection between the Azure Child Site and geographically paired OCI Parent Site. The Parent Site hosts the Oracle control-plane components used to operate and manage the Oracle infrastructure located inside Azure.

The result is an architecture combining Azure-native application proximity and management integration with Oracle's database infrastructure and operational control.

Therefore, the claim that OCI tenancy requirements disappear is false, making A the correct "NOT a benefit" answer.

Study Guide reference: Introduction to Multicloud — Oracle AI Database@Azure architecture and multicloud benefits.

===============

The appropriate architecture is Microsoft Entra ID-to-OCI SAML federation combined with group synchronization/mapping .

SAML provides the authentication component. Administrators authenticate using corporate Microsoft Entra ID identities and obtain federated access to the associated OCI Identity Domain. Oracle's onboarding procedure creates the SAML trust by exchanging metadata between Microsoft Entra ID and OCI.

The provisioning side then maps assigned users and groups from the Microsoft Entra enterprise application into OCI through the configured identity integration. This allows corporate group membership to drive the user's Oracle-side authorization model instead of requiring administrators to create and maintain separate OCI identities individually.

Azure RBAC governs Azure resources but does not by itself create the required OCI IAM group relationships for OCI Console administration. Azure Key Vault controls cryptographic keys and secrets rather than identity federation, while manually creating OCI users directly contradicts the automated identity-lifecycle requirement.

Therefore, A provides both SSO and centralized enterprise identity/group management.

Study Guide reference: Secure Oracle AI Database@Azure — SAML federation, user/group synchronization and OCI IAM.

===============

The purchase process creates an Azure-managed resource group in the background to host the OracleSubscription resource used for Oracle AI Database@Azure billing and subscription integration. Microsoft specifically documents a known issue in which Azure Policy can block this operation.

The system-created managed resource group must initially be located in EastUS , must conform to the provider-defined naming requirement, and is initially created without tags . Consequently, an enterprise policy that allows only West Europe or denies untagged resource groups prevents the OracleSubscription resource from being created. The correct remediation is to exempt the applicable managed resource group from those blocking policies during the purchase process. Microsoft recommends identifying the blocking assignment in the Azure activity log and creating an appropriate, preferably time-bound, Azure Policy exemption.

Administrators cannot solve the problem by manually creating their own resource group in West Europe because this is a provider-managed resource. Billing-address changes likewise do not change the resource provider's required deployment location, and OCI commercial-model changes do not remove the Azure OracleSubscription requirement.

Study Guide reference: Oracle AI Database@Azure Architecture and Onboarding — private offer purchase, OracleSubscription resource, Azure prerequisites and policy considerations.

===============

Oracle's current Oracle AI Database@Azure service matrix explicitly identifies Oracle Exadata Database Service on Exascale Infrastructure as eligible for both Private Offer and Public Offer purchasing models.

A Public Offer uses a pay-as-you-go Azure Marketplace model and allows customers to begin consuming supported services without first negotiating a custom contract with Oracle Sales. A Private Offer is based on negotiated commercial terms and is appropriate for enterprise commitments and customized pricing.

Exadata Database Service on Dedicated Infrastructure is Private Offer eligible but not Public Offer eligible. The same Private-Offer-only restriction applies to Autonomous AI Database on Dedicated Exadata Infrastructure. Oracle Database Autonomous Recovery Service is a database protection service rather than the alternative primary high-performance Exadata database platform requested by the scenario, and it is also listed as Private Offer only.

Exascale therefore combines Exadata database capabilities with a lower-entry, elastic shared-infrastructure model and marketplace flexibility.

Thus, A is correct.

Study Guide reference: Oracle AI Database@Azure Architecture and Onboarding — Azure Marketplace offers and available database services.

===============

The correct staging medium is an Azure Files NFS share . Oracle's ZDM documentation for Autonomous AI Database Serverless on Oracle Database@Azure explicitly supports Azure Files as an NFS file-share source for Data Pump migration staging.

In the logical migration workflow, the NFS share is made accessible to the source database and attached to the Autonomous AI Database target. Data Pump export files can therefore be written to and read from the shared location without being staged in OCI storage. Oracle specifically documents mounting Azure Files NFSv4.x and exposing the path through a database directory for Data Pump operations.

OCI Object Storage and OCI File Storage would place the staging data within OCI, violating the explicit requirement that dump files remain in the Microsoft Azure subscription. Azure Blob Storage can be accessed by Autonomous Database through DBMS_CLOUD for supported object-storage workflows, but the standard ZDM NFS-based Data Pump staging mechanism described in this question uses a mounted filesystem rather than a direct Blob REST API.

Therefore, C is correct.

Study Guide reference: Migrate Oracle Databases to Oracle AI Database@Azure — ZDM, Autonomous AI Database Serverless, Data Pump and Azure Files NFS.

===============

Oracle Exadata Database Service on Exascale Infrastructure uses the Elastic Compute Processing Unit (ECPU) model for allocating database-server processing capacity. Oracle's Exascale technical architecture specifies that a VM requires at least eight enabled ECPUs for initial VM Cluster provisioning and that vertical compute scaling is performed in increments of four ECPUs .

This model provides substantially finer-grained scaling than traditional fixed Exadata infrastructure shapes while maintaining predictable resource allocation across the underlying Exascale compute pool. Administrators can increase enabled ECPUs as workload demand grows, and Exascale also supports reserved ECPU concepts that allow capacity to be physically reserved for future growth. The distinction between enabled and reserved capacity is operationally important: enabled ECPUs are active compute resources used by the VM, whereas reserved capacity can facilitate future scale-up.

An increment of eight would unnecessarily double the required scaling step; one- or two-ECPU increments are not the documented unit for Exadata VM scaling on Exascale Infrastructure. The correct granularity is therefore four ECPUs .

Study Guide reference: Provision and operate Oracle AI Database@Azure Resources — Exadata Database Service on Exascale Infrastructure, ECPU compute model, VM Cluster provisioning and scaling.

===============

The intended topology is Hub-and-Spoke with centralized NVA inspection , making C the correct choice among the available answers. This architecture permits multiple application VNets to communicate privately with the Oracle database while routing inter-VNet traffic through a central firewall or Network Virtual Appliance rather than the public internet.

There is an architectural wording issue worth correcting precisely. Oracle's published reference topology normally places the Azure Firewall/NVA in the Hub VNet while Oracle Database@Azure resides in a separate Data VNet containing the delegated client subnet . Application workloads reside in spoke VNets. The hub then provides centralized routing and inspection between the application spokes and the data VNet.

Therefore, the essential principle embodied by C— Hub-and-Spoke plus centralized NVA inspection —is correct, although the phrase "delegated subnet in the hub VNet" is not how Oracle's canonical diagram lays out the components.

Direct local peering would bypass centralized inspection, while direct Virtual WAN spoke-to-spoke communication violates the same requirement. Data Guard concerns DR, not application traffic inspection.

===============