PadumaTechnology

Services

Architecture modelling

Models that a machine can check, instead of diagrams that quietly go stale

A bounded engagement, or a method your team keeps

What it covers

  • Modelling an existing estate

    Business, information, application and technology, at the levels that matter — one model with one identity space, not a folder of unrelated pictures.

  • A meta-model your models are checked against

    Defining what kinds of thing exist, what may connect to what, and which of those rules a tool can enforce rather than a reviewer having to remember.

  • Solution design that stays connected

    Designing a change against the model you already have, so the target state and the current state are the same artifact seen at two points in time.

  • Getting the practice to stick

    The modelling conventions, the review workflow, and the validation in CI. Without those, a model is a document that ages.

Why most architecture models fail

Not because the notation was wrong. Because nothing stopped them being wrong.

A diagram is checked by whoever reviews the merge request, if anyone does. A model with no enforced meta-model behind it accumulates elements that mean slightly different things to different authors, relationships that were valid in one diagram and nonsense in another, and a gradual drift into a picture that is nobody’s current understanding.

The fix is not a better drawing tool. It is a meta-model — a definition of what kinds of thing exist and what may legitimately connect to what — and something that enforces it automatically.

What we actually mean by governed

Three rules, all of which we apply to our own estate:

An element does not describe itself. Its kind carries its dimension and its layer, declared once in a catalog. An author cannot quietly place the same kind of thing in two different parts of the model.

Conformance is checked by a machine. Structure first, then the semantics the catalog defines: does this kind exist, may these two endpoints be connected this way, are the required attributes present, does this viewpoint still refer to things that exist. A model that fails is rejected, not flagged.

Layout is a projection. How a diagram looks is stored separately from what the model says. Otherwise the picture becomes the truth and the drift starts again.

Tools

The method does not depend on any particular tool, and we will work in whatever you already have. We are informed by ArchiMate’s semantics and UML’s meta-modelling ideas, and we build our own tooling — but we sell the discipline, not a licence, and not a certification.

The honest scope

This is the method underneath our other two lines as much as it is an engagement of its own. If your architecture documentation is out of date and nobody trusts it, that is the problem we are describing. If you need a compliance artifact for an audit, we are probably not the right firm.

What is behind the claims

Each of these names the work that demonstrates it, rather than asking you to take it on trust.

We model our own estate, not a worked example
Our mahjong platform and the Kubernetes cluster it runs on are modelled together as one project — 152 elements and 257 relationships in a single identity space, with written verification notes.
Conformance is machine-checked, not reviewed by eye
A two-phase validator, and a server that validates before it commits and rejects what does not conform. The store cannot hold an invalid model.
Dimension and layer belong to the kind, not to the author
An element cannot declare its own dimension or layer; its kind carries them, once, in a catalog of 36 element kinds and 17 relationship types. It is the rule that stops a model drifting.
We know what a model is worth once it is stale
Every modelling convention we use exists because layout, meaning and identity are kept separate — rendering is a projection, never the source.

What we decline

  • Certification against a modelling standard. We are informed by ArchiMate and UML; we do not sell conformance to either.
  • Tool resale or licence brokerage.
  • Modelling as a deliverable in itself. If nothing downstream consumes it, it will go stale, and we would rather say so.

Next

If any of this describes what you are dealing with, tell us about it.