fredag 2 juli 2010

What is an agile architecture?

Agile is an adjective that mean something along the lines of "quick and well-coordinated in movement", how can we then couple that with architecture?

In my opinion we can’t. Agile architecture is confusing and sounds plain wrong, at least to me who rather think of stability and considered flexibility than quick and nimble.

I can agree that using the phrase “Agile architecture” sets a certain context which can give a reader or listener a notion of what is to be discussed but I still don’t like it. In a similar fashion I don’t like Lean architecture even though I think that the Lean values (the ones that differ from “agile”) fit better with architecture on many accounts. But what is a Lean architecture? Really?

I rather say just architecture. To me the field of software and systems architecture is constantly evolving and the truths I thought correct a couple of years ago are considered mouldy and obsolete today (or maybe I just understand a bigger picture now than back then). It’s important though to point out that what’s evolving is the way we approach architecture, how we work with it etc., whereas the technical aspects don’t evolve as fast; e.g. patterns is still an essential concept, domains as well.

So architecture is architecture but what about Lean and Agile?

Well, I would much rather phrase it along the lines of “Architecture for Lean/Agile organizations/development”. The technical tool-box of architecture doesn’t really differ if you are applying RUP, Scrum or grand scale waterfall, but how you approach it might (and probably will) differ.

I still think architecture is very important in order to reduce cost and make development, maintenance and destruction more efficient so to neglect it would be wrong in my world, but I do think we need to do it a bit differently to achieve better results.

 
  • We need to include more people with different expertise in the beginning to envision where we are heading. 
  • Someone can “own” the architecture as his/hers main concern but we should embrace that everyone contributes to the architecture. 
  • Product management is recognized as a central piece of the puzzle to continuous success and architecture is very tightly coupled to PM.
  • The architecture needs to be documented but we need to improve this area heavily (the old large system architecture descriptions won’t cut it). Documentation should preferably be in such form that it conveys the architecture automatically so reduce the textual descriptions to be the rationale that is needed to understand (and be to the point). Instead document architecture in e.g. code directly where possible, or as structure in an information platform etc. I.e. make the documentation part of the solution.

Architecture is good and well-needed but don’t confuse yourself that there is a hip new cool agile architecture out there.  I have.
 
Have.

tisdag 29 juni 2010

The responsible designer

Some food for thought can be found here.

söndag 27 juni 2010

West side

Spent two days last week in Gothenburg with all the nice colleagues at our main office.

A lot of business development discussions involving our Vinnova projects, product management and software/system architecture for agile or lean organizations.

As usual it was great fun and a learning experience. Thanks all.

Tomas Sandén discussing model-based systems engineering

tisdag 15 juni 2010

New challenge...

... been looking for one, maybe this will do?

SDC

måndag 24 maj 2010

Abstraction is evil


Been a while but attended a very interested seminar last week featuring Jim Coplien (author of the book “Lean Architecture: for Agile Software Development”, among other things) titled “Modern Domain Analysis for Agile System Development”.

I’m not going to try to recap exactly everything Jim said, you’d be better served to listen to him the next time he speak at a venue close to you, but I will dish out some snippets. I don't necessarily agree to everything Jim said but he made some really good points and made me think and reflect which is what I want out of a seminar. As he pointed out this topic is much easier if you are building a web-based application than if you are working with embedded systems, but hey the world is complex. Tough luck.

 
  • Abstraction is evil but compression is good. In Jim’s view all abstraction makes information disappear and we can’t afford to loose information.
  • Architecture is the essence of structure, i.e. the form.
  • Agile architecture is about making the hard decisions early to make the easy decisions easier later. To defer decisions to “the latest responsible moment” is stupid and simply wrong.
  • Agile architecture should provide a stable ground, not necessarily a static ground. It’s about knowing what changes we can ignore.
  • Software architects that don’t write code aren’t software architects. Everyone that writes code is part architect. The whole team should be included doing the architecture; the tester, domain experts, system engineers and programmers.
  • A good architecture should reduce rework and inconsistency.
  • It’s not about no documentation but the documentation should be to the point, only include what matters. The rest of the architecture should be communicated through the code, e.g. abstract base classes.
  • Architecture follows the organization (Conway’s law) and the social connections in the organization (rather than the org. chart, and organization follows location, market etc). Other influences to the architecture is the end-user’s mental model, the shear layers with regards to what the system does and what the system is, and finally the domains as long-term stable concepts.
  • Partition your system to domains using intuition and history. Analyze your domains, analyze commonality. Find your parameters of variation. Use patterns but beware GOF isn’t patterns.
  • DSLs fail.
  • Scrum is very good for systems with end-user interaction and GUI, Scrum is very hard for embedded systems…. basically you aren’t doing Scrum.
  • An incremental approach to architecture leads to poor structure in the long term, you should do it first and then change it when needed.
  • Agile only focus on the revenue but forgets the cost, architecture is a means to reduce cost.  
  • Scrum is waterfall but that doesn't mean it is bad.
  • You work best with “Architecture” influenced by both Scrum and Lean.
  • Scrum is not Lean.
Last but not least. Jim talked around the picture below from an organizational stand point. The picture shows a map of the world from a New Yorkers point of view:


(high-res can be found here)

måndag 26 april 2010

Scrum Scrum Scrum

Scrum can be easy. Scrum can be hard. Scrum can be perfect. Scrum can be so wrong. Scrum can change everything. Scrum can change nothing.

Is it because of Scrum or the way we use it?

Is it just the pessimist stating that Scrum doesn’t work, or doesn’t add any value? Is there any truth to it? Is there truth to the optimist stating that Scrum revolutionized “our” business improving our performance with X percent? Was it because of Scrum or something else?

Personally I think Scrum is one of, or perhaps “the”, best model for team-work out there at the moment. But that’s also it.

I’ve worked for small organizations where Scrum actually might not be needed; it all depends on mindset of the people working together. But I can also see how it would have helped us become more efficient and focused on doing what counts.

For larger organizations though I think that Scrum as the team framework makes, almost, immediate sense. I’ve seen plenty of teams destructively working on unimportant tasks just to generate “heat”. To be stressed out for deliveries even though it is unclear what to deliver and why. To be bogged down by tasks from other teams without the quantity of extra work ever being visible. To be part of a team but completely unaware of what the other guys and girls are filling their days with. What they are actually filling their days with.

But!

To use Scrum in a larger organization is far different from adopting it in the little “SW-shop” with 10 or so developers. During Scrum trainings and seminars I’ve been getting answers implying that as long as you stick to SCRUM you are doing the right thing and that Scrum will point out where your problems are; irrespective if you are 100 or 10 people.

The problem is though that many of these problems can’t be solved by following the rationale of Scrum. Because for large organizations the team-work aspect makes sense but other straight lines in Scrum might need some bending. Or you will at least need to add things to deal with issues you have simply because you are a large amount of people with a lot of interfaces.

Scrum doesn’t have roles (besides the golden 3) and traditionalist state that everyone in the team should able to do any task… is that really practical, cost-efficient, or even possible? How do you put together your teams with this in mind?

Scrum appraises the retrospective but does that really mean you are working towards constant improvement? Does it really?

How do you Scrum when your organization are both assigned with product development and product maintenance? Can you time-box maintenance?

How do you fit testing into it all? Unit-testing sure, how about integration test?

How do you deal with geographical spread?

What do you do if the DoD depend on other teams?

How do you deal with extensive documentation that can be required due to legal or safety aspects?

How long should a sprint be when our cycle-times are long? To short and we add too much overhead, too long and we aren’t Scrumming anymore. Or?

There are a lot of questions around this that I find interesting, for some I think I have the (or at least one) answer and for some I’m still doing some soul-searching.

onsdag 21 april 2010

Leaving the trees behind

Off I go to Gothenburg and the west coast tomorrow; it’s time to talk at the IBC Euroforum conference “Electronics in vehicles”. If anything it will at least be nice to see some old colleagues from Volvo and meet my current Know IT ones.


Besides the very nice assignment at SAAB there’s currently a lot of Know IT business development going on, particularly I’m keeping myself occupied with some ideas around agile development as well as technology management for less mature (in terms of electronics and software) organizations. The space and time for these kinds of discussions and concept evolution is one of things I really like about KIT TM. Oh and not to forget the thread around MDA.

As far as Android goes the time for playing around with Eclipse and the SDK is rather rare but I have been able to play a bit reading sensor data. Besides the serious discussions around apps that I have with a Gbg-buddy of mine I’m keen to try something around augmented reality (after first talking a bit with Torkel and then checking out some demos on youtube).