IBM i High Availability Options: PowerHA, Db2 Mirror, Logical Replication, and Hosted HA

High availability and disaster recovery answer different questions. HA keeps the business running through component failures and planned maintenance, usually with a fast, rehearsed role swap to a live secondary. DR is what happens after you lose the site: recovering somewhere else, with recovery point and time objectives measured in minutes to hours. This guide covers the HA technology choice; our IBM Power disaster recovery guide covers the DR strategy side.

The gap between owning HA and having it is real. Fortra's guidance is that a system that has never completed a full role swap successfully is only set up for DR, not HA, and an August 2025 IT Jungle article reported that many IBM i shops with HA or DR in place had not tested a full role swap in months or years.

PowerHA SystemMirror for i (licensed program 5770-HAS) is IBM's cluster-based HA and DR offering, and it starts with an architectural commitment: application data must live in an independent auxiliary storage pool. The IASP is the unit PowerHA switches or replicates, so a shop that still keeps everything in SYSBAS faces a one-time project of moving libraries and IFS directories into the pool. That work is the toll booth for everything else PowerHA does. Since release 7.5, IBM has sold PowerHA as a single edition, retiring the old Express, Standard, and Enterprise split, and Fortra develops the product on IBM's behalf under a collaboration that began in 2017, when Fortra was still HelpSystems.

Replication comes in two flavors. Geographic mirroring is host-based: IBM i itself mirrors the IASP at the page level over IP, which makes it storage-agnostic (internal disk, external storage, or a mix) and available in synchronous or asynchronous modes. Per Fortra's PowerHA FAQ, synchronous delivers a recovery point of zero at distances up to roughly 300 km, while asynchronous has no distance limit and a recovery point of a few seconds on an unplanned failure; treat the distance figure as guidance and verify against your latency budget. With IBM external storage (DS8000 and FlashSystem lines), PowerHA instead drives Metro Mirror, Global Mirror, FlashCopy, and LUN-level switching, which reassigns the LUNs holding an IASP to another IBM i instance.

Licensing is per processor core with software maintenance, and since IBM's October 2024 announcement, subscription terms from 90 days to five years are available; verify current pricing with IBM or your Business Partner. PowerHA 7.6, which arrived with IBM i 7.6 in April 2025, added Power Virtual Server integrations (IASP LUN-level switching and FlashCopy inside PowerVS, plus Global Replication Service support) along with health dashboards and real-time RPO reporting for geographic mirroring. Expect switchover time in minutes rather than seconds, driven mostly by IASP vary-on and application restart; vary-on scales with the environment, so measure yours.

Logical replication is the incumbent. Products in this family read journal entries, usually through IBM i remote journaling, send them to a target system, and apply them object by object. Because the mechanism is transaction-level rather than block-level, it works across any distance, mixed OS releases, and any combination of storage, including cloud targets such as IBM Power Virtual Server, Skytap on Azure, and provider-hosted LPARs. IT Jungle wrote in 2017 that the vast majority of IBM i HA deployments used logical replication rather than clustering.

Two practical properties define the tradeoff. The target is a live, usable system: vendors note the replica can run tape backups, queries, and reports without touching production, something a block-replicated IASP copy cannot do the same way. In exchange, you own the administrative overhead: choosing which objects replicate, watching apply lag, and running audits and sync checks so you can trust the target when it matters.

If nobody on staff wants to own replication administration, the market will sell you the outcome instead. Connectria offers managed replication, primarily using MIMIX, from customer systems into LPARs in its data centers, with implementation and 24x7 monitoring handled by its staff (it describes itself as the largest IBM Power MSP in North America; a vendor claim). Maxava's Fully Managed Disaster Recovery program, launched in 2023, packages planning, implementation, monitoring, virtual and real role swap testing, and 24x7 crisis support through MSP partners such as Service Express and Meridian IT.

The cloud platforms have moved the same direction. Skytap on Azure announced a turnkey Assure MIMIX integration in October 2023 for cloud-based IBM i workloads, and Fortra documents Robot HA replication into Skytap as well. On IBM's side, PowerHA 7.6 brought IASP LUN-level switching, FlashCopy, and Global Replication Service support inside IBM Power Virtual Server. For how these services are typically packaged, see our overview of managed IBM Power cloud providers. Managed HA costs more per month, but it converts an internal discipline problem, the swap drills nobody schedules, into a contractual obligation.

Whatever you buy, the swap is the product. A replication environment proves nothing until the secondary has actually carried production and handed it back. Fortra's guidance bears repeating: "it's not HA if you've never completed a full role swap successfully. Until you do a role swap, your system is only set up for disaster recovery." The failure mode IT Jungle described in August 2025 is common: HA installed years ago, swaps deferred, runbooks stale, and the people who wrote them gone.

The discipline that works: simulated role swaps for low-risk rehearsal where the product supports them, genuine scheduled swaps so the secondary carries real production periodically, the runbook updated after every exercise, and audits or sync checks running between swaps so drift gets caught early. If a vendor demo dwells on replication speed and never shows the swap tooling, ask why. An HA product whose swaps are never exercised delivers DR at best, at HA prices.