Oracle EBS and Oracle Fusion Cloud to Microsoft Fabric: ERP Migration Guide
Database paths for EBS, API and BICC paths for Fusion Cloud.
Moving Oracle EBS to Microsoft Fabric is less a migration than a choice of integration path. Your ERP keeps running, and Fabric becomes the reporting layer in OneLake and Power BI. The hard part is picking the route that fits your ERP type, latency target and team.
This refreshed guide compares six paths for Oracle EBS and Oracle Fusion Cloud, with the current Microsoft position on mirroring, and gives a checklist for choosing. If you are moving a BI layer too, read our guide on OBIEE to Power BI on Microsoft Fabric.
Architecture Context for Oracle EBS to Microsoft Fabric
Fabric stores data in OneLake as Delta tables, and Power BI reads it through a semantic model, often in Direct Lake mode. Oracle ERP data arrives in a lakehouse or mirrored database, where you model it for reporting. Oracle EBS 12.2 Premier Support currently runs through at least 2037 per Oracle's own page, so most teams are adding analytics alongside EBS, not replacing it. Our guide to migrating to Microsoft Fabric covers the wider programme.
EBS vs Fusion Cloud: Why the Path Differs
The deciding factor is access. With EBS you usually control the Oracle database, so database-level paths are open to you. With Fusion Cloud, Oracle runs the database, so you work through extracts and APIs.
| Question | Oracle EBS | Oracle Fusion Cloud |
|---|---|---|
| Database access | Yes, you manage it | No, SaaS |
| Main routes | Mirroring, GoldenGate, Data Factory | BICC, REST, OTBI |
| Typical latency | Near real time or batch | Scheduled extracts |
| Who owns change tracking | Your database team | Oracle, via the extract design |
Path 1: Native Oracle Mirroring for Oracle EBS to Microsoft Fabric
Fabric lists Oracle among its database mirroring sources, without the preview label that Microsoft still shows against some other sources. Treat that as a signal to confirm, not a guarantee, because Microsoft can change status at any time. Always check the current Oracle mirroring page.
The documented prerequisites matter for EBS. The database needs LogMiner, archive log mode and supplemental logging, and a data gateway connects Fabric to your network. Each mirrored table needs a primary key or unique index, there is a limit of 1,000 tables, and table names must be under 30 characters. EBS has thousands of tables, so mirror the modules you report on, not everything.
Cost is a real benefit. Microsoft says replication compute does not use capacity and that mirroring storage is free up to a limit tied to your capacity size. You still pay for querying. Mirroring needs a running capacity, so size it with our Microsoft Fabric capacity estimator.
Database paths for EBS, API and BICC paths for Fusion Cloud.
Path 2: Oracle GoldenGate Open Mirroring
Open mirroring lets any provider write change data into a landing zone in OneLake. Fabric merges those changes into Delta tables, and Microsoft documents open mirroring as generally available. Microsoft's partner page lists Oracle GoldenGate 23ai, delivered through GoldenGate for Distributed Applications and Analytics, as a partner for Oracle sources.
Choose this path if you already run GoldenGate, need a source that native mirroring does not support, or want change capture without relying on Oracle's native mirroring prerequisites. The trade-off is another licensed product to run. Oracle's original announcement called its integration a public preview, so confirm the current status with Oracle before you plan around it.
Path 3: Fabric Data Factory Oracle Connector Pipeline
The simplest path is a pipeline with the Oracle connector and a copy activity into a lakehouse. It needs the Oracle client for Microsoft tools on the gateway machine, and it is batch, so latency depends on your schedule.
Incremental loading with LAST_UPDATE_DATE
Most EBS tables carry standard WHO columns, including LAST_UPDATE_DATE. A watermark on that column lets a pipeline copy only changed rows. Deletes do not update that column, so reconcile them with a periodic full key comparison. Our guide to upgrading Azure Data Factory to Fabric covers pipeline design in more depth.
Path 4: BICC Export to Storage to OneLake for Fusion Cloud
Oracle Business Intelligence Cloud Connector (BICC) extracts data in bulk and writes it to an external storage location you configure. Oracle's documentation for loading Fusion data into other platforms includes Azure Blob Storage among the options, so you can land extracts in Azure and expose them to Fabric through a OneLake shortcut or a pipeline.
BICC is the standard route for volume because it is built for bulk extraction. Extracts arrive as files with a manifest, so design the load to read the manifest, handle column changes and track which extract you have loaded.
Path 5: Fusion REST API and OTBI
REST APIs and OTBI suit small, targeted pulls, such as reference data or a single subject area, where BICC would be too heavy. They are rate and volume limited, so do not use them for bulk history. Confirm current limits in Oracle's documentation for your release.
Path 6: Partner Tools
Several vendors sell connectors for Oracle ERP. Microsoft's open mirroring partner page lists Qlik Replicate, Striim, Stelo Data and QMigrator as options for Oracle sources, and others offer prebuilt ERP data models. These reduce build effort but add licence cost and another vendor to manage.
- EBS gives you database paths, and Fusion Cloud works through BICC and APIs.
- Check the current status of Oracle mirroring on Microsoft Learn before you plan around it.
- Mirror the modules you report on, not the whole EBS schema.
- Use BICC for Fusion Cloud volume and REST for small pulls.
- Validate against ERP totals and monitor data quality after go-live.
Comparing the Six Oracle EBS to Microsoft Fabric Paths
| Path | Best for | Latency | Effort |
|---|---|---|---|
| 1. Native Oracle mirroring | EBS database | Near real time | Low once prerequisites are met |
| 2. GoldenGate open mirroring | EBS or any GoldenGate source | Near real time | Medium, needs a GoldenGate licence and setup |
| 3. Data Factory Oracle connector | EBS database | Batch, scheduled | Medium, owns incremental logic |
| 4. BICC to storage to OneLake | Fusion Cloud | Scheduled extracts | Medium |
| 5. REST API and OTBI | Fusion Cloud | On demand, small volumes | Medium to high |
| 6. Partner tools | EBS or Fusion Cloud | Varies by tool | Low to medium, adds licence cost |
Pick by latency need and ownership first, and licence cost second. Our Microsoft Fabric migration checklist helps sequence the work, and our Microsoft Fabric migration service covers delivery.
Common Mistakes
- Mirroring all of EBS. Select the modules you report on, because of the table limit and the cost of querying unused data.
- Ignoring primary keys. Tables without a primary key or unique index cannot be mirrored natively.
- Treating Fusion Cloud like a database. There is no database access, so plan around extracts.
- Skipping reconciliation. Compare trial balances and open invoice totals against the ERP before users see a report.
- Leaving data quality until later. Our guide to data quality across platforms shows how to monitor it, and our cost of bad data post explains why it matters.
Oracle EBS to Microsoft Fabric FAQ
Can Microsoft Fabric mirror an Oracle E-Business Suite database?
Fabric lists Oracle under database mirroring, and EBS runs on an Oracle database, so native mirroring is an option when the prerequisites are met. These include LogMiner, archive log mode, supplemental logging, an on-premises data gateway and a primary key or unique index on each table. Check the current Microsoft Learn page for status and limits before you commit.
Does Oracle Fusion Cloud work with database mirroring?
No. Fusion Cloud is SaaS, so you do not have database access. Use BICC bulk extracts for volume, and REST or OTBI for smaller, targeted pulls.
What is open mirroring and how does GoldenGate use it?
Open mirroring lets a partner write change data into a landing zone in OneLake, and Fabric merges it into Delta tables. Oracle GoldenGate for Distributed Applications and Analytics 23ai is listed by Microsoft as a partner that does this for Oracle sources.
Is mirroring free?
Microsoft states that background compute for replication does not consume capacity, and that mirroring storage is free up to one terabyte per capacity unit purchased. Querying the data through SQL, Spark or Power BI is charged at regular rates, and a running capacity is required.
Which path is fastest to start?
For EBS, a Data Factory pipeline with the Oracle connector needs the fewest moving parts, but it is batch. Native mirroring is simpler to run if your database meets the prerequisites. For Fusion Cloud, BICC is the standard route for volume.
How long will Oracle EBS be supported?
Oracle's current page says EBS 12.2 Premier Support runs through at least 2037, and Oracle reviews this annually. Check Oracle's page for the latest, because migration timing should follow your own roadmap, not a support date.
To plan your own route, request a free ERP data assessment, or see our data engineering services. Previous in the series: Xero to Microsoft Fabric. Next: QuickBooks to Microsoft Fabric. See also our Synapse to Fabric migration assistant guide and data maturity scorecard.