Boundaries first, services second
An architecture diagram usually starts with the services: the database here, the application here, the queue there. The result is a map of parts with no edges, where you cannot see what is exposed, what is internal and what belongs to whom.
Designing by zones reverses the order: first you draw the boundaries and then you place each component inside the zone it belongs to. The question stops being 'which services do I use' and becomes 'what can talk to what, and through where'.
Three kinds of zone
Network zones. Where traffic can reach: what is exposed to the internet, what is internal, what is isolated. Every hop between zones goes through a controlled point.
Trust zones. What level of protection each part needs according to the data it handles. Personal or financial data lives in a more restricted zone than a public content service.
Environment zones. Development, testing and production genuinely separated, with no shortcuts between them.
Rules of the method
- Each component lives in a single zone. If it seems to be in two, it is two components.
- Flows between zones are numbered and described: who initiates, over which port or protocol and with what authentication.
- Zones are distinguished by colour, consistently across every diagram in the organisation. The reader learns to read them once.
- The security and observability zone is cross-cutting: logging, monitoring and identity management reach all the others.
- What is pending is drawn differently (with a dashed line, for instance), so the diagram tells the truth about what exists and what does not.
Why it works
A zone diagram answers at a glance the questions that matter in a review: what is exposed, where the sensitive data is, what happens if a zone goes down and what has to be protected first. And because zones are stable, the diagram survives changes of service: you swap a part inside a zone without redrawing everything.
How we apply it: Cloud Architecture.