application architecture · writing · expositional architectures


One of the problems with growing our understanding of software systems is that we don't see enough examples. In many professional disciplines, people learn by looking at what's already been done. Examples serve as inspiration, a source of good ideas, and warnings of difficulties. For a long time it's been much harder to learn about software this way.

I've always enjoyed taking a look at how systems are put together. I hope to get more time to do this and also to share some of the more interesting architectures that I see. I think of these architectures as expositional architectures because they are like show homes or gardens. While they might not be exactly what you would want for your own purpose, they contain aspects that you may want to copy.

For a list of articles of this form, see the expositional architectures tag.

I usually avoid the A-word (architecture) because it's such a slippery term. In this case I'll be following my preferred definition of architecture - "the important stuff (whatever that is)" - from Ralph Johnson. This means I'll talk about what I think is interesting about the system's design, based both on my judgement and on the judgement of the team that's been involved in its development.

I use the term "expositional" to emphasize the fact that these architectures are a source of interesting ideas, and they are not intended to be some kind of "best practice". For a start, I'm very wary of architectures that are set up as some kind of standard, because there are so many variables to pay attention to when building a concrete system. For example, many people stress the importance of a scalable architecture (by which they usually mean the ability to handle large amounts of load). Yet many useful systems are internal systems that never have a high load, so should be designed to support a different set of drivers.

It's often useful to highlight factors that haven't been successful. We don't just learn by looking at things that worked well, things that have been unsuccessful can often be a good guide for avoiding a tempting path. And it's common for some architectural feature to be really liked by some team members and hated by others. By understanding what drives the affection and dislike, you can get an impression on whether it would fit in with your architectural aesthetic.

I won't be making expositional architecture posts very frequently, since they do take a lot of time and effort to write. It takes a while to understand a system well enough to get a feel for how it works and where the interesting bits are. It also requires time and effort from the development team to explain things and verify I haven't got any sticks by the wrong end.

I expect most of the architecture that I'll be describing will come from ThoughtWorks projects, as they are easier for me to get to. However this isn't an absolute rule and I'll happily talk about non-TW projects should something catch my eye when I have the time to work on it [1].

Any expositional architecture I write about will be based on at least one real system. I prefer something that's been in production for a year or so, because many things that look good in development turn out differently once you've been live for a while. But that is a preference rather than an absolute rule.

When describing the architecture I'll give the architecture a name to act as a reference, but usually this name will be one made up for my description, since system names are otherwise too tied to the client's context. Similarly I describe systems using a vocabulary consistent with the rest of what I write about here, which may not be the actual terms used by the development team.


1: Indeed the first expositional architecture I've written about was LMAX - which wasn't a ThoughtWorks effort. (Although it was an ex-ThoughtWorker working on it who was my point of contact.)