SATURN 2013 Software Architecture Conference

The Software Engineering Institute (SEI) Architecture Technology User Network (SATURN) Conference took place at the Marriott City Center in Minneapolis, Minnesota from April 29 to May 3, 2013.  I attended the keynotes and sessions on May 1 and 2.  It was a mix of practicing software architects, developers, managers, and academics.

The keynotes were longer form, lasting about an hour with everybody in the main hall.  The session blocks were 90 minutes that included three shorter presentations.  This was in effect about 20-25 minutes for each, with time for questions at the end.  There were three separate session tracks available to attend each block.

The day 1 opening keynote was about how the “Credit Suisse Information Bus” took about 15 years to realize their service-oriented architecture efforts fully. It was an interesting case study in how long some efforts take, how to integrate new systems via acquisitions and how a few of the services have the most reuse with the long tail dropping off fast.  The afternoon keynote was about working at Wordpress and how extreme they are on self-direction.  Anyone can check anything into production, but they only hire all-star talent.  This may work for a 50 to 150 person company, but how high does it scale?  However, this was a good counter perspective to the large-systems, big-up-front design methodology.

I attended a set of Modeling and Documentation sessions in the morning block, where I was reminded of ISO/IEC/IEEE 42010 and the more formal descriptions of software architecture.  These sessions were more academic and formal.  In the afternoon block, I  attended Agile II.  Some good phrases were “document significant decisions vs. low-level details” and the concept of ”risk storming.”  A session that references some SEI studies hit home, as the top highlighted issues with agile projects were “features over stability”, “slow business decisions” and “external dependency management.”  However, the largest group in their study was a >30 person integrated team (includes developers, managers, business, architect, etc), so an order of magnitude less than our WestlawNext agile implementation.  I attended an open space/breakout session at the end of the day for a good ad-hoc discussion.

On day 2, Mary Poppendieck delivered the morning keynote, which was more a psychological presentation on the rational vs. intuitive mind how that impacts lean and agile projects.  Mary is always an entertaining speaker.  The closing keynote was my favorite, where Philippe Kruchten was speaking on “Games Architects Play: On Reasoning Fallacies, Cognitive Biases, and Politics.”  He has a set of patterns of behaviors and biases, including: golden hammer; “yes, but”; teacher’s pet; “let’s have a vote”; and “obviously.”  The point was knowing these patterns is a useful tool to avoid their effects.

I attended a morning set of sessions called “Fusion Methods” where I heard about Risk- and Cost-Driven Architecture (RCDA) and reminded me every architect should have a backlog that they are working against.  Another session talked about agile planning and used a “map vs. a GPS” metaphor and the balance at looking at the map up front for full directions vs. use a GPS and only knowing the “next step” which sometimes results in getting lost because of bad information and hearing the dreaded “recalculating”.  I jumped over to see a talk on the next generation of web architecture style, referred to as Service-Oriented Front-End Architecture (SOFEA).  This was a very good talk for 20 minutes but it is basically what we designed for ADC - (I should have created a fancy acronym).  In the afternoon, I saw some “Architectural Evaluation” sessions.  In case studies of in-the-field evaluations, teams usually have “good people” but the requirements are missing or not documented and had to be reconstructed for the onsite evaluations.  Unfortunately, this reminded me of some projects I have been on where the next feature was prioritized over tech debt, including good tests and documentation.  When I learned how Raytheon has to test missile guidance systems, I am glad we have web applications we can hotfix.

More information can be found at the SATURN 2013 conference site, including a download all the SATURN 2013 presentations.