Why organisations still run them
Nobody keeps a mainframe out of nostalgia. They keep it because the cost, risk and time of moving forty years of accumulated business rules somewhere else is enormous, and because for a certain shape of workload the platform is genuinely very good.
The honest answer has four parts
- The workload fits. High-volume, short, repetitive transactions against a shared database — card authorisations, policy quotes, benefit payments, airline reservations — are exactly what the platform was built for.
- The rules are in the code. Decades of regulation, exceptions, edge cases and 'we do it this way for the 1987 acquisition' live inside COBOL and PL/I programs. Often no separate specification exists. The code *is* the specification.
- The cost of being wrong is asymmetric. A payments platform that is 99.9% correct is a scandal. Migration projects introduce risk in exchange for savings, and for some workloads that trade is not attractive.
- It is not standing still. Modern mainframes run Linux, containers, Java, Python and REST APIs alongside the traditional stack, so 'move off' and 'modernise' are not the same decision.
Where you will find them
The part people get wrong
'Mainframe' and 'legacy' get used as synonyms, and that muddles two separate things. The hardware and operating system are current products under active development. A particular *application* on top of them may well be legacy — poorly documented, unowned, hard to change. Those are different problems with different fixes, and conflating them makes for bad conversations with the rest of the business.
Common mistakes
Both are marketing positions. The useful question at any given site is: which workloads genuinely fit this platform, and which are here only through inertia?
The hard parts are usually undocumented business rules, data integrity during cutover, and regulatory sign-off — not rewriting code.
What you will see at work
- You will meet applications with no owner and no documentation. Reading code to recover business rules is a normal, respected activity here, not a failure.
- Expect a modernisation programme running in parallel with day-to-day support. Both are real work; be careful about which budget you are charging to.
- Cost conversations are constant. Reducing CPU consumption in a hot batch job is genuinely valued work.
Key terms
Check your understanding.
Take this lesson's quiz and save your progress. Free.