Mainframe costs are often described as high, opaque, and uncontrollable. Two of those three are sometimes true. The third is not.

The difficulty is not that mainframe costs are inherently incomprehensible. It is that most IT leaders have never had a proper introduction to how they work and live with an incorrect preconception about these costs. That gap leads to bad decisions, both in managing existing mainframe environments and in evaluating alternatives.

Where the money goes

Running z/OS applications on IBM Z falls into three cost categories: hardware, software, and management – the split surprises most people.

Many outsiders think the mainframe hardware is the expensive part. But if you look at the real figures, software can account for up to two-thirds of total mainframe cost. Hardware and management each account for about one-sixth. If you are trying to control mainframe costs, you are primarily managing a software problem. Therefore, understanding how software costs work on the mainframe is the most important thing you can do.

The number that drives your software bill

Most mainframe software, I am focusing on z/OS here, is licensed under a model called MLC (Monthly License Charge). The monthly amount is variable, calculated based on how much CPU capacity you used that month, measured in MSUs (Millions of Service Units).

More specifically, your software bill is based on your peak four-hour rolling average of CPU usage in a given month. Not your peak second. Not your peak hour. Your peak four-hour average – the 4HRA.

These are some abbreviations thrown at you, but understanding the 4HRA is critical. Your workloads can spike as high as they need to without affecting your software bill, as long as those spikes do not push up the four-hour average. This distinction matters enormously, and it is one that many mainframe users, including specialists, find confusing.

When the cost model can become a problem

The MLC model was designed for stable, predictable workloads. Modern workloads are not always that.

When a z/OS application grows significantly and increases CPU consumption, it can push up the 4HRA, and, with it, the cost of other software products on the same platform, regardless of whether those products were involved in the growth. The shared platform creates a shared cost structure.

I have seen organizations invest in a new customer-facing application, deliver a genuine business win, and then discover that the resulting transaction volume pushed their 4HRA so high that the business case collapsed. Not because the mainframe was too expensive. Because the cost model was not understood before the decision was made.

This is just one problem you face when you are not well-informed about your IT costs.

Is this unique to the mainframe? It is not. Growing applications on other platforms can also cause unexpected cost impacts elsewhere. I have seen a growing web application cause a very significant increase in database license costs. And there are enough stories nowadays about unexpected costs in cloud environments. It is not about the platform. It is about understanding the cost structure. That is what enables informed decisions.

The comparison problem

When organizations compare mainframe costs to cloud or x86 alternatives, they almost always compare the wrong things.

A mainframe is not a single server. It is an integrated platform that combines computing, storage, I/O processing, redundancy, and security in a compact configuration. A fair comparison includes all the x86 servers, networking, storage, switches, the required energy consumption, and management overhead needed to achieve equivalent capability, including running all your development, test, acceptance, and production environments, which on a mainframe typically share one or two machines.

Make that comparison carefully, and the cost picture looks very different.

The good news

New license models exist for new workloads. If you are adding a new application to the mainframe or expanding existing ones, your software vendor will offer more flexible and competitive licensing options, often far below your traditional MLC rates.

The catch: you have to ask, and know what to ask. These models are not automatically applied. Organizations that do not have this conversation pay more than they need to, for longer than necessary.

What this means in practice

Mainframe costs are complex. But they are manageable if you understand the model. The organizations that struggle most with mainframe costs are not those with the highest bills. They are those who do not understand what drives the bill, and therefore cannot control it.

This chapter in my book Don’t Be Afraid of the Mainframe comprehensively covers this topic: a full cost breakdown, the MSU and 4HRA model, the difference between hardware and software MSUs, new license models and their pitfalls, and a framework for making informed cost comparisons.

---

Available on Amazon, Barnes & Noble, and other bookstores.

Links to order the book, available in epub and print.
Amazon: https://lnkd.in/e6SNBy9f Barnes & Noble: https://lnkd.in/eigqF9Js 24bookprint: https://lnkd.in/eNTur2AU, coming soon on Bol.com, Apple Books, Google Play, and other bookstores.