Deciding the architecture

Agile project starts by gathering an Agile team, deciding on a project and specific deliverables to be built first, and setting up the needed infrastructure. 

But before the team goes about creating the infrastructure, several questions about the system architecture need to be addressed.  What infrastructure will be needed, and what skills the team must have, greatly depends on the proposed system architecture and technology stack.

Who decides system architecture is the “elephant in the room” issue of Agile projects.  Many smaller design decisions will happily “emerge” as the team goes through the daily work of building the project iteration after iteration. Still, there is a number of large-scale, long-lasting choices about system architecture, that must be decided at the start of the project, when relatively little is known and the team has the least amount of experience on the project.

Who should be making these decisions? The typical answer appears to be The Architect, someone high up in the ivory tower, someone who makes no mistakes by producing little to none tangible, and thus imperfect, results. Sometimes that person is very good, and in other cases they just get by.  As it turns out, making the very best initial architectural decisions for a given project is not particularly important for the project's success, and making passable architectural decisions is fairly easy and fool-proof. 

List of common architectural patterns is very small, and almost all projects fit nicely into that tiny set.  Even if an inappropriate pattern is picked, the team will be fighting, and winning, over the inappropriate pattern to create a working implementation.  If the system becomes successful despite poorly designed architecture, and fighting the system setup becomes a serious obstacle to evolving the product, it will get re-written with a less inappropriate architecture shortly.

P.S. 

Sydney Opera House is one of the most beautiful and distinctive buildings in the world. Its architecture went through a dozen iterations of design, and the finished structure exhibits terrible acoustics - a big problem for an opera house and performance venue.  The building took 10 years longer than planned, and went 1,357% over budget. Of course, the project was, and still is, a resounding success.