GoldenGate licensing, where topology is the invoice
GoldenGate licenses per processor on every server that runs it, source and target alike, on the same counting rules as the database. The surprise is symmetry: both ends of every flow carry the full count, so the server topology, not the data volume, decides the bill, and one initialization parameter decides what is in scope.
Prepared by Redress Compliance · August 6, 2026 · Oracle licensing advisory. Based on the Oracle practice engagement record 2024 to 2026.
Executive summary
GoldenGate lists at $17,500 per processor and $350 per Named User Plus, metered on the database's own counting rules, core factor included on premises, and the crossover between the metrics sits at exactly 50 named users per licensable processor.
The number that surprises buyers is not the rate but the symmetry: source and target each carry the full processor count, so a modest replication mesh multiplies into a large license position through topology alone.
Scope is decided by one initialization parameter.
ENABLE_GOLDENGATE_REPLICATION set to true puts a database in the count, and Oracle reads the parameter as covering third party replication tools as well, so a Qlik or Striim deployment can create a GoldenGate license claim on databases that never ran Oracle's own product.
The parameter is what the audit scripts read, and it is what your inventory has to read first.
The Foundation Suite is genuine arbitrage. At $7,500 per processor it contains Veridata and the Management Pack, which list separately at $33,500 combined.
On a 32 processor estate the difference is $832,000 of list and $183,040 a year of support, for the same components, purchased in a different order.
The hub question is worth 5x. Concentrating GoldenGate on a hub server does not remove the source and target databases from the count unless your ordering document explicitly says so: on the worked estate in this guide, the two readings priced at $140,000 and $700,000.
The clause, not the architecture, decides which one you owe.
The metric, database rules applied to replication
GoldenGate meters on the same processor basis as the Oracle Database: physical cores times the core factor on premises, vCPU counting in the authorized clouds, Named User Plus as the alternative where populations are small and countable.
Every server running the software counts, and running includes extraction at the source, delivery at the target, and anything in between that executes GoldenGate processes.
| Component | List price | The note |
|---|---|---|
| GoldenGate | $17,500 per processor, $350 per Named User Plus | Both ends of every flow, at the database's counting rules |
| Foundation Suite | $7,500 per processor | Contains Veridata and the Management Pack, $33,500 of separate list, the arbitrage |
| Veridata, standalone | Part of the $33,500 combined separate list | Data comparison tooling most estates buy twice by accident |
| GoldenGate Free | $0, perpetual | Fenced at 20 GB per database, one deployment, no support, no mixing with licensed instances |
Symmetry is the cost model. A three source, one target consolidation flow licenses four servers. A bidirectional pair licenses both directions once, on both machines.
The design conversation that treats GoldenGate as a per link cost has already miscounted; the unit is the server, and every server it touches is the estate.
The parameter, and the third party surprise
ENABLE_GOLDENGATE_REPLICATION is the scope switch: set true, the database is in the count.
Oracle's reading extends the parameter's coverage to third party replication and CDC tools that require it, which is how estates running Qlik Replicate or Striim against Oracle databases acquire GoldenGate exposure without ever installing GoldenGate.
The parameter is precisely what the measurement scripts collect, which makes it precisely what your own quarterly inventory reads first.
The defense is the standard Oracle discipline: know which databases carry the parameter, why, and under what entitlement, before any measurement asks. Databases that enabled it for a migration that ended, or for a tool that left, are cleanup, not liability, if you find them first.
The full third party analysis sits in the GoldenGate for non Oracle databases guide.
The GoldenGate licensing playbook
The topology counting worksheet, the hub clause language, the Foundation Suite arbitrage math, and the parameter inventory method for the whole replication estate.
Get the white paper →The hub question, one clause worth five times the bill
The architectural instinct is to concentrate GoldenGate on hub servers, fewer machines running the software, smaller count.
The contractual reality is that the hub design only prices as a hub if the ordering document says the licensed servers are the hubs alone; absent that language, Oracle's position reaches the source and target databases the hub serves.
On the worked estate behind this guide, the hub only reading priced at $140,000 and the full reach reading at $700,000, the same architecture, five times apart.
The lesson generalizes: GoldenGate savings are drafted, not designed. The topology creates the opportunity, and the ordering document either captures it in writing or donates it back at the first audit.
Any hub consolidation business case that has not been read against the actual contract language is a proposal, not a saving.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
The cloud edge, where the core factor stops
On AWS and Azure the core factor table does not apply and the vCPU rule takes over, two vCPUs per processor license with hyperthreading on, exactly as for the database, so the same topology counts differently across the cloud boundary.
A flow that was core factored to a modest count on premises can double on migration, and instance families running one thread per core double it again. The counting mechanics are the same ones worked in the WebLogic on AWS guide and the Oracle on Azure analysis, applied to every server in the flow.
GoldenGate Free deserves its honest place at the edge of the estate: a real perpetual license, fenced at 20 GB per database, one deployment, no Oracle Support, and no mixing with licensed instances.
It fits dev, test, and small isolated flows, and it fits nothing else, which is exactly the fence the terms draw. The full price family context sits in the technology price list guide.
The patterns from the engagement file
Across the Oracle practice's GoldenGate reviews, the findings cluster in three places, none of them throughput:
The gap between hub only and full reach contract readings on the same topology, decided by ordering document language.
Foundation Suite at $7,500 against $33,500 of separately purchased components, plus $183K a year of support on the difference.
The third recurring finding is the parameter estate: databases carrying ENABLE_GOLDENGATE_REPLICATION for tools and migrations long gone, each one a claim waiting for a measurement.
The quarterly parameter inventory, joined to the entitlement position, is the cheapest control in the Oracle estate relative to what it prevents.
Your first five moves
- Inventory the parameter across every Oracle database, quarterly: which carry ENABLE_GOLDENGATE_REPLICATION, why, and under what entitlement.
- Map the topology to the count: every server running GoldenGate processes, both ends of every flow, at the correct counting rule per environment.
- Read the hub language in your ordering document before believing any hub consolidation saving. Draft it explicitly at the next order.
- Buy the Foundation Suite arbitrage deliberately, and stop paying separate support on components the suite contains.
- Fence the Free tier honestly, dev, test, and sub 20 GB isolated flows, and keep it unmixed with licensed instances. The Oracle practice and license management services run the estate with you.
Frequently asked questions
How is Oracle GoldenGate licensed?
Per processor on every server that runs the software, source and target alike, at $17,500 per processor list or $350 per Named User Plus, on the database's own counting rules, core factor on premises, vCPU counting in authorized clouds.
The crossover between the two metrics is exactly 50 named users per licensable processor.
Do both the source and target need GoldenGate licenses?
Yes. Every server running GoldenGate processes carries the full count, which is why topology rather than data volume decides the bill. A consolidation flow from three sources to one target licenses four servers, and a bidirectional pair licenses both machines.
Can third party replication tools trigger GoldenGate licensing?
Yes. Oracle reads the ENABLE_GOLDENGATE_REPLICATION initialization parameter as the scope switch, and third party CDC tools like Qlik Replicate and Striim require it on Oracle sources.
Databases carrying the parameter are what the measurement scripts collect, whether or not GoldenGate itself was ever installed.
Does a GoldenGate hub reduce the license count?
Only if the ordering document says so. Absent explicit hub language, Oracle's position reaches the source and target databases the hub serves: on our worked estate the two readings priced at $140,000 and $700,000 for the same architecture. The clause, not the topology, captures the saving.
What is the GoldenGate Foundation Suite and why does it matter?
A $7,500 per processor bundle containing Veridata and the Management Pack, which list separately at $33,500 combined. On 32 processors the difference is $832,000 of list plus $183,040 a year of support, for the same components.
Estates that bought the pieces separately are paying the arbitrage every renewal.
Is GoldenGate Free suitable for production?
Within its fence: it is a genuine perpetual license limited to 20 GB per database, one deployment, with no Oracle Support and no mixing with licensed instances. It fits development, test, and small isolated flows, and using it beyond the fence converts a free tier into an audit finding.
