① Licensing · IFL
An IFL takes a Linux workload off z/OS MSU, and with it the MLC charges that track the MSU peak. That is real money saved. But the software stack on the IFL is licensed per core, so the saving on one side can be partly handed back on the other. The buyer side question is never whether IFLs help, but where the per core licensing lands once the workload moves.
Off MSU, onto cores.
An Integrated Facility for Linux is a specialty processor on IBM Z and LinuxONE dedicated to running Linux and z/VM, where one IFL equals one core. The cost significance is simple: work on an IFL does not consume z/OS MSU and does not attract z/OS Monthly License Charge software charges. That is why eligible Linux workloads are placed on IFLs rather than general processors. The IFL is one of the specialty engines, alongside zIIP and zAAP, that exist precisely to keep certain work out of the MSU based bill.
What changes is the metric, not the existence of cost. Software on IFLs, the Linux distribution subscription, middleware, and ISV products, is licensed per IFL or per core rather than by MSU. Much IBM middleware uses a Processor Value Unit model, and ISV products set their own per core or per IFL terms. The workload leaves the MSU world and enters the per core world, and the two have to be compared, not assumed.
What the cores actually cost.
Under a Processor Value Unit model, the entitlement for a middleware product is the PVU per core rating multiplied by the number of cores it runs on. Take a middleware product on an estate of 6 IFLs with an illustrative PVU per core rating of 70, licensed at full capacity:
| Input | Value | Calculation |
|---|---|---|
| IFL cores | 6 | Physical cores available to the program |
| PVU per core rating | 70 | Illustrative, set by IBM per processor |
| Full capacity entitlement | 420 PVU | 6 × 70 |
| Sub-capacity (4 cores virtualized) | 280 PVU | 4 × 70 |
| Entitlement avoided by sub-capacity | 140 PVU | The reporting requirement earns this |
Illustrative PVU rating for arithmetic only. Real per core ratings are set by IBM per processor model and vary; ISV products use their own metrics.
Two lessons sit in that table. First, IFL software cost scales with cores, so an oversized IFL pool licensed at full capacity pays for cores the workload may not need. Second, sub-capacity (also called virtualization capacity) licensing lets you license only the cores made available through virtualization, here 4 instead of 6, but only if you maintain the tracking and reporting the vendor requires. Skip the reporting and full capacity applies by default, which quietly hands back the saving. The per core world has its own version of the same discipline the MSU world demands.
Four ways to count a core.
There is no single IFL licensing metric. The stack on a Linux LPAR can mix several, and knowing which applies to each product is half the cost control.
| Model | Unit | Where it shows up |
|---|---|---|
| Per IFL / per core | Each IFL core | Linux subscriptions, simple per core ISV terms |
| Processor Value Unit (PVU) | Cores × PVU rating | Much IBM middleware on Linux on Z |
| Full capacity | All available cores | Default where sub-capacity is not tracked |
| Sub-capacity / virtualization capacity | Virtualized cores only | Where reporting is maintained |
ISV products on Linux on Z frequently set bespoke per core or per IFL terms, so the contract governs, not a single industry rule.
The trap is mixing the saving on one metric with a cost on another. Moving a workload to IFLs to escape MSU is sound, but if the workload drags a heavy middleware stack licensed per core, the per core bill can erode much of the MSU saving. The decision is never IFL or not in the abstract. It is this specific workload, with this specific stack, priced both ways, with sub-capacity claimed where the reporting supports it. For the broader Linux on Z and LinuxONE pricing picture, see software pricing on LinuxONE.
How to optimize it.
Levers on IFL licensing
Across 500+ engagements and $180M+ of negotiated mainframe spend, IFLs are a genuine cost lever when the per core stack is priced as carefully as the MSU saving that justified the move. The estates that disappointed were the ones that treated escaping MSU as the end of the analysis rather than the start of a second one. The core based bill is real, and it is just as movable as the MSU based one.
A specialty processor on IBM Z and LinuxONE dedicated to running Linux and z/VM, where one IFL equals one core. Work on an IFL does not consume z/OS MSU and does not attract z/OS MLC charges, which is the main reason Linux workloads are placed on IFLs rather than general processors.
Not by MSU. Linux subscriptions and middleware on IFLs are typically licensed per IFL or per core, frequently through a Processor Value Unit model where entitlement equals the PVU per core rating times the cores, with full capacity or sub-capacity options. ISV products on Linux on Z set their own per core or per IFL metrics.
They can, because moving an eligible workload to Linux on an IFL removes it from z/OS MSU and the MLC charges that track the peak. But the saving is not automatic: the stack on the IFL is licensed per core, so a workload needing many IFLs and heavy middleware can carry significant per core licensing. Run the comparison, do not assume it.
Under full capacity you license all IFL cores physically available to the program. Under sub-capacity, or virtualization capacity, you license only the cores made available through virtualization, which can be fewer. Sub-capacity requires the tracking and reporting the vendor specifies; without it, full capacity applies.
Audit notice or renewal under 18 months out? We mobilize within 48 hours.
Get expert help →