As Agile becomes more accepted, concerns from architecture groups are increasing. Traditional ways that architects engage with development groups conflict with Agile methods. This talk describes ways that Agile methods can benefit architects, addresses concerns architects express about agile, and proposes ways that architects and agile development teams can become allies.
Martin Fowler is an author, speaker, consultant and general loud-mouth on software development. Rebecca Parsons has more than 20 years of application development experience in industries ranging from telecommunications to emergent internet services.
QCon is a conference that is organized by the community, for the community.The result is a high quality conference experience where a tremendous amount of attention and investment has gone into having the best content on the most important topics presented by the leaders in our community. QCon is designed with the technical depth and enterprise focus of interest to technical team leads, architects, and project managers.
What I seem to find out is that Agile goes hand in hand with large and distributed systems architecture. By nature these systems already change a lot. We've been trying to reach patterns since the beggining of ages that allow for easy maintenance, no (or low) impact on performance, and most of all, accelerated development. This requires design both at a high level as well as a low level.
Yes, we're agile. However, we cannot drop the high level design up front. We need to have a (not very clear) vision of what we are going to need. And agile copes with that. I found out, when moving to OutSystems, that you are provided with the most relevant patterns from scratch. The very nature of the platform extension mechanisms won't let you easily fail. It also provides you a kind of Domain Specific Language approach from the beggining, that will force you to create (and all developers to follow) well defined guidelines. All the assynchronous design patterns are also easily mapped to the patform potentials in timers. When coping with external systems or a corporate company architecture definition there's not much to trouble us as well. We're either at SOA (which OutSystems fits like a ring) or some other kind of integration pattern, for which you know you'll have an integration layer only for transforming/transporting/auditing/etc., that will be either implemented out of the box, or via a simple extension. All this facilities reduces the effort in defining the high level architecture, assuming you're confortable with the usual design patterns. This way you don't have to go very deep on the initial architecture definition. Just focus on business components/modules decoupling.
I am now in a software factory with a very large architecture basis. The only thing required is that, whenever a new requirement appears, you do an overall architecture impact analysis. If you have the above topics considered, you end up in having a new component. If the business requirement changes (and you'll adopt it "immediately" because you are an agilist), it will probably fit in its own business component, or have relatively low impact on the surroundings.
What about when there's a requirement that reaches all levels? If you've decoupling from the start the Core entities (business domain core - the ones you usually specify the DSL for) from the business components/processes on top of them and also from the integration layers, there's no big issue. You'll probably end up in refactoring some core processes to decouple even more, so that the new requirement fits in an isolated way.
This proved me that agile will make you good architectural choices. And if you know from the start you're going agile, the architecture will come out much stronger and maintainable, just because of the fact that you're thinking, from the beggining of ages it will be ment to change.
OutSystems Solution Delivery team