Legacy systems: strangle, do not replace

Enterprise Technology Arjun Mehta · 1 min read

Big-bang replacements are the most expensive way to modernise. Routing traffic one path at a time is slower on paper and far cheaper in practice.

Everyone knows the story: three years, a nine-figure budget, and a system that does less than the one it replaced. The alternative is not glamorous, but it works.

Put a seam in front of it

A routing layer between your users and the old system costs weeks, not years. From that point on, every route can move independently.

Move the cheapest risky thing first

Not the biggest, not the easiest — the one that teaches you the most about the seam while still being safe to get wrong.

Run both until the numbers agree

Parallel running is tedious and it is what makes the cutover boring. Reconcile daily, and do not switch anything off until the difference is zero.

Decommission on purpose

An old system nobody has switched off is still costing you licence fees and audit scope. Plan the archive before you plan the migration.

It takes longer to describe than a replacement. It takes considerably less time to survive.

The short version

0

unplanned outages on our migrations

25+

systems moved this way

daily

reconciliation during parallel run

Questions this raises

Slightly more in engineering, dramatically less in risk and downtime. The projects that blow budgets are the ones that go dark for eighteen months.

Then the seam goes at the database or the screen. It is uglier and it still works — we have done both.

Want to talk this through?

We are happy to have the conversation whether or not it turns into work.

Contact us