Explainer · Licensing concepts

On LinuxONE the IFL core is the meter. The MSU never enters the room.

Software on IBM LinuxONE is licensed per IFL core, not by MSU. That puts the whole cost model on a different axis from z/OS Monthly License Charge software, with its own levers and its own audit traps. Here is how core based pricing works, why LinuxONE work stays off the z/OS bill, and a worked per core example.

A different processor, a different meter.

IBM LinuxONE runs Linux and z/VM workloads on the Integrated Facility for Linux, the IFL processor. Software on it, the Linux distributions, middleware, container platforms, and databases, is licensed primarily per core or per virtual processor core, not on the MSU metric that drives z/OS Monthly License Charge software. The IFL carries no MSU rating and does not raise the MSU rating of z/OS workloads on the same machine. So the LinuxONE software bill runs on a separate axis from the traditional mainframe bill, governed by core counts and virtualization rules rather than by a Rolling 4-Hour Average peak.

This separation is the structural reason organizations consolidate distributed Linux onto LinuxONE or onto IFLs in a shared frame: the Linux workload is licensed per core and stays entirely off the z/OS MSU meter, while the underlying engine is one of the specialty engines that sit outside MLC. The levers, and the audit exposure, move from peak management to core counting. This explainer pairs with our IBM audit defense, where core based products are a standard area of scrutiny.

LinuxONE against z/OS MLC

Why the two cost models do not share a lever

Dimensionz/OS MLC softwareLinuxONE software
Processor General purpose CP IFL, dedicated to Linux and z/VM
Metric MSU on the R4HA peak Core or virtual processor core
Reporting Monthly SCRT submission Core counts and virtualization records
Main lever Peak shaping, capping, zIIP offload Core counting, virtual core sub-capacity, consolidation
Effect on the other bill Drives the z/OS bill Does not touch the z/OS MSU meter
Main audit trap Missed SCRT defaulting to full capacity All cores counted where sub-capacity not configured

A worked per core example

The core count is the whole question.

Take a container platform on LinuxONE running across a cluster with 16 IFL cores, but the workload is sized to use 6 virtual processor cores. If the product permits virtual core sub-capacity counting and the configuration is documented, you license 6 cores. If sub-capacity counting is not applied, or not evidenced, the same workload can be billed against all 16 physical cores in the host. The work is identical; only the counting basis differs. The illustration indexes the per core charge at 100 units to make the gap visible without quoting vendor pricing.

Physical IFL cores in the host16 cores
Virtual processor cores the workload uses6 cores
Indexed per core charge100 units / core
Billed at full physical cores (no sub-capacity)1,600 units
Billed at virtual cores (sub-capacity applied)600 units
Avoided by counting cores correctly1,000 units · 63%

The units are an index, not vendor pricing, and the gap depends on the product's rules. The structural point holds: on LinuxONE the core count is the bill, so virtual core sub-capacity counting, where the product allows it, is the equivalent of getting sub-capacity right on z/OS. The audit trap is being billed against every physical core because the sub-capacity configuration was never set or never documented.

Controlling LinuxONE software cost

№ 01

Count cores against the rules

Know whether each product allows virtual core sub-capacity counting and configure for it where it does. The difference between physical and virtual core counting is the LinuxONE equivalent of full versus sub-capacity on z/OS.

The core count is the bill; count it deliberately.

№ 02

Cap and pin the workload

In a virtualized cluster, affinity and capping keep a workload from exposing more cores than intended. Without them a mobile workload can drift across cores and inflate the licensable count.

A workload that roams is a workload that bills wide.

№ 03

Consolidate distributed Linux

Linux running on distributed servers and licensed per socket or core can often consolidate onto IFLs at a lower aggregate core count, and stays entirely off the z/OS MSU meter. Model the consolidation as a licensing move, not just a hardware one.

Fewer cores, and none of them on the z/OS bill.

№ 04

Document the entitlement

Keep records of the core counts you are entitled to license and the sub-capacity configuration behind them. On a per core metric, the evidence of how many cores you license against is what turns an audit query into a non event.

The core count you cannot prove is the one they will charge.

Frequently asked questions

Q1

How is LinuxONE software licensed?

Primarily per IFL core or virtual processor core, not by MSU. Linux distributions, middleware and container platforms are counted per core across the cluster. The IFL carries no MSU rating and does not raise any z/OS MSU rating.

Q2

Does it affect my z/OS MSU bill?

No. IFL work does not count toward the general purpose MSU that drives z/OS MLC software, and the IFL does not add to the MSU rating. That separation is why Linux consolidates onto LinuxONE without touching the z/OS meter.

Q3

Per core versus virtual core?

Per core counts physical or logical cores; virtual processor core counts the cores assigned to the workload and can count fractionally across a cluster. For products like OpenShift on IBM Z the subscription is core based, with one IFL typically mapping to one core subscription.

Q4

What are the audit traps?

Core counting and virtualization scope. Per core software can be billed against all cores in a host unless sub-capacity virtual core counting is allowed and configured. Set affinity and capping, and keep evidence of the core counts you license against.

Related: licensing concepts hub · specialty engines explained · sub-capacity vs full capacity · IBM licensing hub · IBM audit defense

Audit notice or renewal under 18 months out? We mobilize within 48 hours.

Get expert help

Billed against every core in the host? Sub-capacity core counting is the lever.

Get expert help