Legacy Architecture
A legacy architecture is a computer system that has grown over years or decades and that a company continues to operate even though the underlying technology is outdated. It usually works reliably, but is difficult to change, extend, or replace.
Large companies often use computer programs that were written twenty or thirty years ago. Such old, but still running systems are called legacy architecture. The English word “legacy” means inheritance or heritage. What is meant is the basic structure of a piece of software: which building blocks exist and how they work together. This structure fits the requirements of back then, not those of today. Nevertheless, the system keeps running because important daily work depends on it.
Why banks are sitting on 50-year-old code
The best-known example is banks. Their core systems, which manage accounts and process transfers, often date back to the 1970s and 1980s. They were written in COBOL, a programming language that hardly anyone learns anymore today. According to estimates, billions of lines of COBOL code are still running worldwide in the financial sector.
The reason for holding on to them is not convenience. These systems have been tested over decades and make almost no errors. A rebuild, on the other hand, costs hundreds of millions of euros and takes years. Several banks have started such projects and then abandoned them again. A mistake during the transition would mean that customers could no longer access their money.
For business news, the topic is relevant because legacy architectures tie up money. Studies estimate that large corporations spend more than half their IT budget just on maintaining old systems. This money is then unavailable for new products. Whoever modernizes faster has an advantage over the competition.
How good software becomes a legacy burden
No system is planned to become legacy. At the start there is a clean solution to a specific problem. Then, year after year, new requirements are added: an online feature, a legal requirement, a special rule for a major client. Each change is bolted onto the existing structure because a rebuild would be too expensive. After twenty years, the whole thing resembles a house with thirty extensions.
A second problem is lost knowledge. The people who built the system have long since retired. Documentation is missing or outdated. No one can say for certain anymore why a particular part of the program exists. So no one dares to delete it either.
Also typical is the tight interlocking of the building blocks. Experts speak of a monolith: all functions are contained in a single large program. If you change something in one place, something else can break in a completely different place. Modern systems are therefore broken down into small, independent services, so-called microservices.
Where the term appears in news and everyday life
You encounter legacy architectures more often than you might think. When the railway website reports a disruption, when a government agency only accepts applications on paper, or when an online shop only processes orders overnight, an old system is often behind it. The error messages at airports during system outages usually have this cause as well.
In corporate news, the term comes up in the context of modernization projects. Companies then announce that they are moving their legacy systems to the cloud, meaning onto rented computers from external providers. Such announcements almost always come with high costs and long timelines. Investors pay attention to this because delayed IT projects can trigger profit warnings.
Recently, AI has started playing a role here. Language models can read old COBOL code and explain what it does. IBM and other providers sell tools that automatically translate such code into modern languages. A common misconception, however, is that this solves the problem. Translated code has the same convoluted structure as before, just in a new language.