Architecture Iterations

I am attending a "transition to agile project management workshop" by Johanna Rothman. It seems most of the attendees are from large corporations -- Cisco, CA, SAP and the like -- which in my opinion is a good sign. I see a lot of benefits in agile values and it good to see interest from large companies.

Many of the projects me and the other attendees seem to have are rather large and include all sort of challenges for introducing agility; e.g., few participants mentioned off-shoring some of the development, another mentioned hardware integration, etc.

To me, that is just a reminder why JEDUF is important. I find that in projects that are large or overly complex "sacrificing" one, two, or even three iterations for handling technical risks and forming a candidate architecture goes a long way (and I don't care if this makes my project not agile. I am fine if it is pliant, lagum or what-not). Some things are harder to add or change at later iteration. prototyping the user interface (or more importantly the overall user experience) is one example, others are related to availability, security and basically many of the system's quality attributes. This doesn't mean that you "implement by architecture" as Johanna mentions in a recent blog. You still want to implement full features going through the different layers and/or components of the architecture since evaluation in code is probably the best way to prove your candidate architecture actually fits.

Posted by Arnon Rotem-Gal-Oz at 03:20 PM  Permalink