Just Enough Architecture

How much architecture is just enough? Is a common question I get which I will try to address in this blog.

Determining the right level of architecture design can be challenging; its very easy to get caught in a loop of over engineering for some people, and hard to know where expectations start and end for others. So what is the right level?

The High Level Design

The high level design template I created for work runs to 42 pages. Its an extremely long document that was built out of necessity. It has a lot of helping text and example views. It attempts to do a couple of things:

  • Provide a document structure that is ISO42010 compliant
  • Provide a document structure that provides Business, Data, Application, Technology Descriptions, and can be aligned to TOGAF
  • Provide extra spaces for common integrations and considerations we have within our own organization.

It aims to be usable by architects of very different levels of experience. When new architects look at it – it can be scary. Even though I wrote it, I really do not like it. I have discussed my thoughts on documenting architecture in document templates in the past, and I would much rather focus on the views we need to see to address the concerns of the stakeholders than trying to crank something into a rigid template.

Architecture View Points

I previously showed in this blog this diagram of the different types of viewpoints I like to consider as a minimum in a design, and my HLD template covers these, however there are times this can be considered overkill.

HLD Viewpoints
Viewpoint Structures

We determine standard viewpoints in architecture to make sure the concerns of our stakeholders are addressed and communicated. For architecture to make sense, it has to provide value. There are cases where it may not make sense to create all the views shown above.

An Example – Transforming On Premise Services to the Cloud

For example, lets say we have a team that’s been supporting on premise versions of Exchange & SharePoint, and we are creating an architecture as part of a proposition to extend the scope of services they provide to include Office 365 cloud based solutions.

If we are going to consume office 365 as a cloud service it may not make sense to create technology views, given that the technology layer is by Microsoft. It doesn’t provide value to anyone for me to rehash Microsoft’s standard material in ArchiMate.

In the same way, if we are going to be reusing an existing team, with existing processes, it doesn’t make sense to redefine all the processes – we simply do the definition for the things we need to specifically provide. If creating a specific view doesn’t make sense, there’s probably no need to create it. There are cases when its not obvious to your audience why a view doesn’t make sense – and you can document that.

If it doesn’t make sense to do a technology view, you may want to mention the reasons why in the service realization view for example. Sometimes if you just miss something, it will lead to an unnecessary volley of questions that could have been avoided if you had just written a sentence.

We are telling stories

We are trying to tell a story to our stakeholders, and explain an often complex set of elements and their connections. We are also trying to address stakeholder concerns and provide consistent easy to understand architectures. We should structure information in a way that is easy to consume. Sometimes that means changing the way a standard template is, or building custom views to clearly indicate our subject to our stakeholders. I spoke about telling stories in a previous blog on Improving ArchiMate Modelling Quality.

Documenting what we need to

Last week I was designing a solution based on Microsoft Sentinel in Azure. The challenge here is in the way that solutions are built – I could go into great detail about how data sources connect into the analytics engine, how rules are created and processed, and how they trigger playbooks. This is all detailed at some length in various places in Microsoft, and to define everything doesn’t make much sense.

But I did have to define a basic skeleton of azure architecture at an application level, because without it, it wouldn’t be possible to understand my architecture at all in a connected way without experience on Sentinel – which I cannot assume my audience has.

Not missing the essentials

There are some things you always need for an architecture design to be complete – and forgive me if you have heard this before:

An architecture should define and show how it meets the concerns (or requirements) of different stakeholders, and identify and manage related risks.

Without that an architecture is not complete, and will raise problems and there will be consequences in the future. The same thing is true if you do not design to meet the needs of all your stakeholders.

I speak often about things such as the architecture aspects, and architecture principles, and security non functional requirements – the point of these things is always the same – to ensure that we have captured the needs of our stakeholders.

Summing it up

Anyone who focuses on just creating a document is missing the point of architecture practice – its not providing its full value if its only a documentation exercise – its the process of creating a design that mitigates risks and meets requirements that is really providing the benefit – documenting is a natural part of that process, but at the end of the day, personally I do not care how big an architecture is, as long as it provides value to its identified stakeholders. I previously wrote a blog 8 reasons not to leave architecture to the end of a project – which could have also been titled 8 reasons not to think of architecture as a documentation exercise. Architects need to understand our standards, and be able to maintain a balance between what is too much and what is too little. Above all, architects must think.