Guide

Mainframe modernisation approaches, compared.

Rehost, replatform, refactor, rewrite and AI-assisted migration: what each changes, what each costs, and how to know it worked.

In this guide6 parts
  1. 01Rehost
  2. 02Replatform
  3. 03Refactor (automated translation)
  4. 04Rewrite
  5. 05AI-assisted refactoring
  6. 06How to choose, and how to know it worked

In short

  • There are four classic mainframe modernisation approaches: rehost, replatform, refactor and rewrite, plus newer AI-assisted refactoring.
  • They trade speed and risk against how much of the system you actually change.
  • Whichever you choose, the deciding question is the same: can you prove the new system behaves like the old one?

Rehost

Move the same COBOL to new hardware or cloud, using an emulation layer for the mainframe environment. It is the fastest route and changes the least behaviour, but the code is still COBOL and the skills problem remains.

Replatform

Recompile the COBOL on a different runtime and swap supporting pieces, such as the database. You keep the code but leave the vendor. Behaviour mostly carries over; testing focuses on the platform differences.

Refactor (automated translation)

Translate COBOL into a modern language such as Java or C#. This is where the long-term value is: the system becomes code your current team can read, change and hire for. The catch is proof: refactored code has to be shown to behave exactly like the original, which is why verification has to be built in from the first change.

Rewrite

Rebuild the system from its requirements. It offers the most freedom and the most risk, because the requirements usually have to be recovered from the code first through business rule extraction.

AI-assisted refactoring

Large language models now translate and explain COBOL quickly. The model alone is not the safeguard: the reliable pattern combines program analysis to give the model structure, execution-based testing to check the output, and human review before anything ships.

How to choose, and how to know it worked

Choose by what you need to change: the platform, the language or the design. Then plan verification from day one, not at cutover. Replay-based migration testing checks each step against recorded legacy behaviour, and a parallel run confirms the whole system before the switch.

Questions

What are the main mainframe modernisation approaches?

Rehost, replatform, refactor and rewrite, with AI-assisted refactoring as a newer variant of refactoring. They differ in how much of the platform, language and design changes.

Which mainframe modernisation approach is lowest risk?

The one you can verify. Rehosting changes the least on day one but keeps the COBOL and the skills problem. Refactoring removes that risk for good, as long as every step is checked against the legacy system's real behaviour, which is what Montora is built to do.

Is AI good enough to migrate COBOL on its own?

Not on its own. AI speeds up translation and explanation, but the result must be checked against the legacy system's real behaviour and approved by people who understand it.

Sources

Keep reading

See it on your own code.

Montora is taking on a small number of design partners: banks, insurers and other teams with COBOL they can't afford to get wrong.