Migration Tooling on OCI: ZDM, Data Pump, and GoldenGate Costs

Migration tooling conversations suffer from a peculiar inversion: the tools are discussed as if they were the decision, when they are the consequence of one. The real decision is how long the system can be down, priced honestly per hour. Once that number exists, the tooling selects itself, and most of the time it selects something free. This article walks the OCI migration toolbox in cost order, from the included workhorses to the licensed replication tier, and gives the framework for matching tool to workload without paying for capability the outage window never asked for.

This article is part of our series on what an OCI migration really costs, and it pairs directly with pricing downtime, which supplies the number that drives every choice below.

The free tier: what ships with the database

Data Pump is the default logical migration tool and costs nothing beyond the compute it runs on. It exports schemas or full databases and imports them to any OCI database service, handles version upgrades and platform changes in the same move, and its limitation is the one that defines the whole category: the data is a snapshot, so the source must stay quiet from export to import or the delta is lost. For databases up to a few terabytes with a tolerable weekend window, Data Pump remains the most cost effective migration method in existence, and quotes that reach past it without a downtime justification deserve a question.

RMAN backup and restore moves the database physically: a backup lands in OCI Object Storage and restores onto the target service. It is byte faithful, well understood by every Oracle DBA alive, and free. Cross platform moves need transportable mechanics, and the window is the backup transfer plus restore plus archive log catch up, which for large databases over modest bandwidth becomes the bottleneck. The hidden cost is purely infrastructural: Object Storage for the backup and the network path it travels.

Data Guard deserves more credit as a migration tool than it gets. A standby built in OCI, kept in sync with redo, becomes the production database at switchover time, shrinking the window to minutes. For Enterprise Edition customers the capability is part of the license they already own, making it free replication for the subset of estates entitled to it. The costs are the standby's infrastructure during the sync period and the engineering to build and validate it, which is real but modest for teams that run Data Guard anyway.

The free orchestration tier

Zero Downtime Migration is Oracle's free orchestration layer over the methods above. In physical mode it automates the Data Guard pattern: builds the standby in OCI, manages the sync, runs the switchover. In logical mode it orchestrates Data Pump with GoldenGate on top for change capture. ZDM itself costs nothing; it runs on a small host and turns a manual, error prone sequence into a repeatable, resumable workflow. The budget line it creates is effort, not license: installing it, learning its job model, and rehearsing, typically a few days of skilled work that repays itself by the second database. For estates moving many databases, ZDM converts migration from artisanal work into a production line, which is exactly the learning curve economics that wave planning monetizes.

OCI Database Migration is the managed service wrapping of the same machinery, driven from the OCI console. The service itself is free for offline migrations and prices modestly for online ones; what it mainly buys is accessibility, putting validated migration workflows in front of teams without deep ZDM experience. Its constraint is coverage: source versions, target services, and network paths must fit its supported matrix, and estates at the edges of the matrix fall back to running ZDM directly.

The licensed tier: GoldenGate

GoldenGate is the tool people mean when they say zero downtime, and it is where money enters the conversation. It captures changes from the running source and applies them continuously to the OCI target, shrinking cutover to a connection repoint, and enabling the fallback replication that derisks the nights nobody talks about. It is also genuinely costly, in three possible shapes. Classic on premises licensing prices per processor on terms that make sense only if GoldenGate has a life after the migration. The OCI GoldenGate service prices by OCPU hour consumed, which for a migration measured in weeks is dramatically cheaper than perpetual licensing, and it scales to zero when idle. And some Oracle cloud programs have bundled GoldenGate entitlement for migration use, which is worth asking about before paying list for anything, the kind of contract question an independent licensing review answers definitively.

The budgeting rule we apply: GoldenGate enters the picture when the downtime price per hour, multiplied by the window the free tier would need, exceeds the all in replication cost, service consumption, the dual running it implies, and the skilled configuration days. That calculation lands in GoldenGate's favor for genuine tier one systems and against it for almost everything else, which is why an estate wide GoldenGate quote is a red flag worth interrogating line by line. The fallback window question also belongs here: keeping reverse replication running for a week after cutover extends the consumption but buys a rollback path measured in minutes, which for the systems that justify GoldenGate at all is usually worth the extension. The wider parallel environment costs of that period are the subject of budgeting dual running.

The comparison in one table

The costs that hide around the tools

Whatever the tool, four costs travel with it and rarely make the quote. Rehearsal infrastructure: every serious cutover rehearses at least once, and the rehearsal needs a target environment that exists for days or weeks, billable OCI consumption that belongs in the migration budget, not the run budget. Network: the data must physically cross, and bandwidth determines whether a 10 TB transfer takes hours or days; FastConnect provisioned for the migration often stays for the estate, but its lead time belongs on the critical path. Ingress is free on OCI, but the source side of the path, especially egress from another cloud, can carry real charges. Validation effort: row counts, object comparisons, application smoke tests, the unglamorous hours that decide whether the cutover gate opens, scale with database complexity, not size, and are chronically underquoted. And skilled time itself: the free tools are free the way a piano is free once you can play it, which is the actual argument for bringing in a team that has played it across 500+ engagements, on a fixed project fee through our OCI implementation practice with 20+ years of combined Oracle experience and 24/7/365 coverage through cutover windows. Estates with a long migration runway sometimes prefer the managed monthly retainer shape instead, and where the move is done but the spend looks wrong, the optimization model charges only on verified savings.

Beyond the database: the application tier toolbox

Database tooling dominates the conversation because databases dominate the risk, but the application tier has its own toolbox and its own cost profile. For straightforward server moves, OCI offers free paths: boot volume and image exports, the Oracle Cloud Migrations service for rehosting virtual machines in coordinated waves, and for VMware committed estates, Oracle Cloud VMware Solution, which moves workloads with their hypervisor assumptions intact at the price of dedicated capacity. File payloads travel by rsync or object storage staging for ordinary volumes, and by the Data Transfer Appliance when the dataset is large enough that shipping disks beats sending bits over any affordable line. None of these carry license costs; all of them carry the same satellite costs as the database tier, bandwidth, rehearsal, and validation hours.

The application families have their own accelerators, covered in their own articles in this series: PeopleSoft Cloud Manager and JD Edwards One Click Provisioning industrialize environment builds the way ZDM industrializes database moves, and the budget logic is identical, a few days of setup that converts bespoke effort into catalog operations. The general rule across both tiers is that Oracle has already built free automation for the move it wants you to make; the quote should show the effort to adopt that automation, not line items for capabilities the platform includes.

Estate scale tilts the choices too. A two database estate should not buy orchestration sophistication it will use twice; a forty database estate cannot afford to skip it. The crossover sits lower than most teams expect, around the third or fourth repetition of any move type, which is why the tooling decision belongs in the wave plan rather than being made per workload on the night.

Version coverage is the last gate before any tool earns its place on the plan. Every tool in the table carries a support matrix, source database versions, target services, platform combinations, and the matrix is least forgiving exactly where old estates need it most. A database stranded on a version from two support cycles ago may fall outside ZDM's comfortable range, force an interim upgrade, or push the move down to plain Data Pump regardless of what the downtime price would prefer. Check the matrices against the real estate inventory during assessment, not during cutover planning, and treat any required interim upgrade as its own budget line with its own testing, because that is what it is. The pattern repeats often enough to state as a rule: the older the source, the fewer the tooling options, and the more the free tier ends up being the only tier, which is one more argument for pricing the windows honestly rather than promising near zero downtime an unsupported tool cannot deliver.

A framework for choosing per workload

The tooling decision is the rare corner of migration budgeting where the cheap answer and the right answer usually coincide. Let the downtime price pick the tier, let ZDM industrialize whatever the tier is, buy GoldenGate by the OCPU hour only where the arithmetic demands it, and spend the money saved on the rehearsals and validation that actually decide whether cutover night ends well. Database moves and their cost drivers in the wider sense are covered in Oracle database migration to OCI, and the estate level picture in the pillar that anchors this series.