IBM i Migrate While Active: what it costs and what it won't do

IBM shipped three resilience-flavored things in the summer Technology Refreshes: updates to IBM i Migrate While Active, a batch of BRMS improvements, and IBM i support for the TS7785 virtual tape library. Two of those are resilience features. The headline one is a migration tool, and the IBM engineer who built it has said on the record that it is not a disaster recovery product. If you read the TR coverage and came away thinking you now have a cheap warm standby for $2,000 per core, that is the wrong conclusion, and it is an expensive one to reach in the middle of an outage.

What Migrate While Active actually is

Migrate While Active is a repurposed piece of Db2 Mirror. Where Db2 Mirror uses RDMA over Converged Ethernet to keep two systems in lockstep within a 10 kilometer radius, MWA drops the exotic networking and moves database and application changes asynchronously over ordinary TCP/IP, at any distance. It builds a copy of your production partition on a target system, keeps that copy synchronized while production keeps running, and then lets you cut over on a schedule you choose.

That is genuinely useful. It is also narrower than it sounds.

MWA does not use journaling. It uses Db2 Mirror’s asynchronous replication built on tracking lists, which is far lighter on storage than a journal-based approach because it does not have to record every change in sequential order. The tradeoff is the one that matters: the data on the target is incomplete until the replication session ends. You cannot pull the plug on production mid-migration and boot the target.

There is also still downtime, twice. Once at the beginning to create the checkpoint that starts change tracking, and once at the end to sync source and target. IBM’s claim is that a migration that used to take days or weeks of downtime now takes minutes or hours. That is a real improvement and it is not the same as zero.

One practical detail that catches people: you cannot use your existing monthly full system save as the initial save. IBM needs a checkpoint to know exactly when change tracking starts, so the save has to be taken as part of the process.

What it costs, and how the licensing works

This is the part that no one puts in one place, so here it is.

The Db2 Mirror base product, 5770-DBM option 5050, is licensed at no charge. On top of that you pick option 1, Continuous Availability (5101), or option 2, Migrate While Active (5102). Picking MWA does not require you to license full Db2 Mirror.

MWA is licensed per core on a subscription. At general availability IBM quoted $700 per core for a 90-day subscription and $2,000 per core for a year, with a 28-day test period available first. IBM i customers can also apply $2,000 in earned cloud credits toward it. Those figures date to the December 2024 GA and IBM has not publicly restated them since, so treat them as the shape of the pricing rather than a quote, and get your own from your rep before you budget.

For a small partition this is cheap relative to what a migration weekend costs in overtime and risk. For a large one, per-core pricing scales the way per-core pricing always does. Run the numbers before you assume it is a rounding error.

What the summer TRs added

Three things, and the first one is the most useful news in the whole release.

You can now perform an OS upgrade as part of the migration. This is the change that turns MWA from a cloud onboarding tool into something a shop with no cloud plans should care about. Kris Whitney framed it directly: “As part of that migration, whether I’m moving to the cloud, moving to a new piece of hardware, or maybe I’m not even going to move to a new piece of hardware, I’m moving to a new operating system. How do I use some of this technology to reduce my downtime?” If your 7.5 to 7.6 upgrade window is the thing standing between you and the features in this year’s Technology Refresh, this is a lever you did not have last year.

You can seed the target from external storage. In addition to save/restore and manual methods, you can now create the initial target copy using FlashCopy or Global Copy. Whitney expects FlashCopy to be the most popular option, and if you already have IBM external storage under the box, it is the fastest way to stand up a target without a long save.

Partition mirroring now works on 7.6 and 7.5. IBM announced partition mirroring in July 2025 with 7.4 support only. With TR2 and TR8 it works on 7.6 and 7.5 as well. This is the method worth understanding, because it replicates below the technology-independent machine interface. No initial outage on the source, no database locks or conflicts with running production, plus a test mode that lets you exercise the target fully before you commit.

There are also new estimation and planning enhancements in the GUI, which sounds like filler until you have tried to explain to a business owner how long a cutover will actually take.

Which of the four methods you should pick

MWA now documents four ways to establish and sync the target. None of them removes the planning work that makes or breaks a cutover, and what actually happens when IBM i shops migrate has more to do with data quality and interface inventory than with replication mechanics. Here is the short version.

Partition mirroring if your environment supports it. No source outage, no lock contention, test mode, and clean progress reporting in the Db2 Mirror GUI. This is the default answer.

External storage if you already run IBM external storage and can use FlashCopy or Global Copy. Fast initial copy, and you can upgrade the OS on the target while production keeps running on the source.

Assisted save and restore if you have neither of the above. You take an initial outage to create install media and a user data save, then the guided process handles the rest. Fine, just budget for that first outage.

Manual if you have a specific reason the assisted path does not fit, for example you want to use GO SAVE option 21 or QSRMIGRATE for the full system save, or you want to mix storage replication with save/restore. More control, more rope.

On sizing: Whitney’s guidance is that data estates of 2 to 3 TB or less are usually fine on MWA’s native replication depending on your bandwidth. Above 10 TB, a virtual tape library’s compression and dedupe start to matter enough that you should plan the initial copy around one.

The DR trap

Say it once more, because the framing of the coverage invites the mistake. Whitney, on the record: “This is not a DR solution; this is a migration solution. We’re not trying to keep everything in lockstep while we’re migrating it.”

The temptation is obvious. You have a synchronized target sitting there, running, for the price of a subscription. It looks like a standby. It is not one, because the replicated data is incomplete until the session ends, and a partial system is not a recoverable system. If you want continuous availability, that is Db2 Mirror option 5101, RoCE adapters, distance limits, and a different budget. If you want DR, that is PowerHA, a replication vendor, or a well-tested backup and restore process.

Worth noting alongside this: if you have turned on the strictest form of the new bring-your-own-key encryption in 7.6, your recovery process now depends on supplying a Save/Restore key at IPL. Test that path on the target before you need it, not after.

BRMS 2026.2.0: web GUI only

Every BRMS change in this release lives in the web interface. If your operators work in the green screen, this TR gave them nothing. Set expectations accordingly. Worth knowing before you plan around any of it: that web interface only ships in the paid subscription version, and BRMS licensing quietly changed from a free, bundled product to a $400 per core per year subscription back in 2024, a fact most shops still running the old version have not caught up to.

What you get in the GUI: filtering the BRMS log by severe messages, meaning message IDs critical enough to fail a control group, and filtering by messages containing “objects not saved” text. Object-level and file-level detail views in Backup History Management, with the ability to restore at object level when object detail was included in the save, and to restore link data at file level on the same condition. Backup list management picked up add, copy, remove, and edit for list entries. The SQL views got a documentation link.

The item actually worth acting on is administrative: a new role-based usage type, *SYSOPR. It grants read and view access to media and backup information, allows the user to run backup and restore operations, and blocks change, copy, and remove actions unless the user owns the BRMS item. “Can run the backup, cannot rewrite the backup” is a separation of duties most shops have wanted for years and have been faking with authority lists. If you have auditors, this one is worth the PTF on its own.

The PTFs are SJ09069 for 7.4, SJ09070 for 7.5, and SJ09071 for 7.6.

The TS7785: a mainframe VTL learns IBM i

IBM brought the TS7700 family down from System Z to distributed platforms, and the summer TRs added IBM i support for the TS7785. This is the first time IBM i can attach to a TS7700-series device at all.

The specs: Power9 processor, 1.57 PB raw capacity per VTL, up to 11.77 PB compressed, emulating a physical LTO 5 drive-based TS4500. Up to eight 32 Gb/sec Fibre Channel connectors give you backup speeds up to 4 GB/sec. It integrates with BRMS and IBM Storage Protect, and it works either natively or through VIOS. It handles in-line compression and encryption, geo-dispersed replication, and policy-based tiering.

The part your team has not operated before is the GRID architecture. Multiple TS7785 units link together, replicate between themselves, and your tapes are reachable from any point in the grid. Whitney’s own framing was that it is “slightly different than how virtual tape libraries work today in the industry,” which is IBM’s way of saying your existing VTL mental model does not transfer cleanly.

IBM does not publish list pricing for the TS7785, so this is a quote-only conversation, and it is one where you have competition. LaserVault ViTL, FalconStor, and DSI have all been selling VTLs into IBM i shops for years and know the workload. Get a number from IBM and a number from at least one of them.

Bottom line

The most useful thing in this release is not the flashy one. It is the quiet expansion of Migrate While Active to cover in-place OS upgrades, which turns a cloud onboarding product into a downtime-reduction tool for shops that were never going to the cloud. If a 7.6 upgrade has been sitting on your roadmap because you cannot get a long enough window, price out MWA for the duration of the project and see whether a 90-day subscription beats another year of waiting.

The BRMS changes are minor unless you use the web GUI, in which case the *SYSOPR role is a real win. The TS7785 is a serious box worth evaluating if you are already replacing tape infrastructure, and premature if you are not.

And whatever you do, do not file Migrate While Active under disaster recovery. IBM has been clear about that, even if the coverage has not been.

If you are working through a wider IBM i modernization plan and want the AI-assisted side of it documented in one place, the IBM i AI Field Guide covers the prompts and workflows we use for legacy code analysis and documentation.

Frequently Asked Questions

How much does IBM i Migrate While Active cost?

IBM licenses it per core as a subscription. At general availability IBM quoted $700 per core for 90 days and $2,000 per core for one year, plus a 28-day test period. The Db2 Mirror base product (5770-DBM 5050) is free, and Migrate While Active is option 5102 on top of it, so you do not need a full Db2 Mirror license.

Can I use Migrate While Active as a disaster recovery solution?

No. Kris Whitney, who leads the Db2 Mirror development team and is now IBM i chief architect, said plainly that it is a migration solution and not a DR solution. The replicated data on the target is incomplete until you end the replication session, so a target you have not cut over to is not a system you can recover onto.

What changed in BRMS in IBM i 7.6 TR2 and 7.5 TR8?

All of it landed in the BRMS web GUI, version 2026.2.0. You get severity and “objects not saved” filtering on the BRMS log, object-level and file-level views in Backup History Management with matching restores, backup list add/copy/remove/edit, and a new *SYSOPR role-based usage type. The PTFs are SJ09069 for 7.4, SJ09070 for 7.5, and SJ09071 for 7.6.

Does the IBM TS7785 virtual tape library work with IBM i?

Yes, as of the summer 2026 Technology Refreshes. It is the first time IBM i can attach to a TS7700-series device, which was previously a System Z product. It supports both native and VIOS client connections, and integrates with BRMS and IBM Storage Protect.