Sunday, April 13, 2008

Agility, Biology and SOA


 

I have just finished writing a new chapter for the third edition of my book « Entreprise Architecture & Business Process Management » (which is only published in French, unfortunately). This new chapter deals with "SOA and Agility", two buzzwords of these IT times.

A part of this chapter deals with the concept of "sustainable architecture" following what I wrote in my previous post. I have proposed to paraphrase the definition from the Brundtland report:

"Sustainable development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs."

into a definition of that spells out the "sustainable development" of the IT architecture:

"Sustainable architecture is the Entreprise Architecture that enables the development of today's IT services without compromising the ability to develop tomorrow's services because of an ever-increasing complexity or the induced scarcity of resources (money or skills)."

To gain this sustainability, I have been advocating for quite some while for "fractal" architectures (hence, "fractal" methods). I mean methods that can be broken into pieces and applied differently (and at different rates) to different situations. This fractal view applies to middleware, to enterprise architecture, but also to data models or to process models. The information system (as a whole) needs to grow (as a living object) according to different rates of pressure that are applied to its different parts. The CEISAR architecture documents make a very elegant and clear point about the value that exists when breaking long "end-to-end" business processes into smaller, more manageable pieces.

The major part of this new chapter is devoted to agility (mostly because I hear too much nonsense about solving the agility issues with technology). To give a short summary, I see four forces that contribute to agility:

  • Anticipation: being agile is implementing changes rapidly enough to meet TTM (time-to-market) constraints. A smart way to think about it is to start early. It is not a "solution in itself" (since many changes may not be anticipated) but it is definitely part of the "whole package" since many "smart decisions" (the one that will increase agility) take time to get implemented.
  • Flexibility: the technical ability to cover a large span of business situations and processes with the existing set of IT services and capabilities.
  • Leanness: the process of applying changes to the information system (for instance, developing a new IT project) needs to be as lean as possible (cf. my other blog for more details, although I will most certainly be back on this topic). Lean obviously means short from an organizational perspective (the longer the chain between the developer and the problem owner, the less agility is observed). The ultimate goal is to see IT project disappear. The "absolutely agile" IT department is the one who is not needed to let the users adapt their information system to the business changes. Although it may sound silly, it is actually a sound goal, but one that requires a lot of anticipation.
  • Skills: one needs to be competent to be agile! This also sounds dumb, but actually it is a rather profound observation, which is backed by observation. The fear of failure and the pressure in today's modern organizations means that any lack of full understanding or full confidence will translate into multiple redundant checks, back-and-forth, "safety precautions", etc. Another dimension is the ability to profit from the opportunity provided with the constant progress of technologies (from new IHM techniques (Ajax, mash-up, etc.) to new middleware tools, including more radical options to leverage the value in SaaS – software as a service).

Notice that, except for the second one, these are organizational traits and not technical ones. This, by the way, is also true of being able to leverage the value of SaaS. The software services provided by SaaS are not especially great. Using them is smart from a business and an organizational perspective, and requires "business agility/flexibility" even more than technical skills or abilities.

As I wrote earlier, SOA's main contribution is to provide a framework for sharing and reusing services. It also helps to achieve agility, but more as a second-order consequence rather than a direct cause-effect path, and not without efforts and dedication. The first-order benefits of reuse and share are cost and complexity reduction. On the other hand, as I have already explained earlier, SOA is foremost a governance principle, before being an architecture framework. If the stakeholders do not play "the game", the existence of a beautiful ESB (enterprise service bus) will not provide as much benefits as expected. As a governance principle, SOA sets principles and rules that must be followed.

If this sounds very different from a crude idea of what a "biology-inspired" IS could be … it is indeed! The strength of biomimecry as a design principle is found in what nature does very well:

  • Flexibility/ adaptability (to new situations)
  • Reliability / Survival ability

One thing that nature does not do so well is the "economy principle", the "simplicity of design". When biologists study mechanisms in living objects, they are often surprised by the level of redundancy, by the complexity of the processes (which have been obtained through evolution). By the way, when we look at legacy code, we often get the same impression J

There are two consequences:

  • Biomimecry is not enough as a design principle. The economic reality of business requires implementing a "principle of simplicity" and a constant strive to reduce costs. I advocate for a "fractal architecture", which may sound like an oxymoron to some, as a weakened form of the necessary dictatorship of principles and rules. Rules are necessary to produce simplicity.
  • For all the good reasons mentioned in earlier posts, it is a good idea to borrow from Nature to design systems which are at the same time: complex, reliable, flexible (doing any of the two is OK, doing the three is a challenge). However, this will come at a cost (mostly redundancy). To justify the effort, one must look at the "big picture" and consider full "life cycles" of IT services and take multiple business scenarios into account. This means that the projects that will bring biomimecry into the IS architecture will not have a simple ROI justification, but rather a sophisticated one. This is discussed in my book, which should be made available in English (at last) in a few months.

Nature is organized around a simple principle that says that evolution cannot be stopped, which also applies to information systems. Hence the only mechanism to cope with complexity is to clean, to remove unnecessary things. Nature does great job of cleaning (and recycling) unnecessary things in living organisms. This is another aspect which may be borrowed, especially when managing a SOA. A key issue is to master the complexity (and the sheer number) of services. The book will mention a few tools and techniques to manage this catalog of shared services, but the first principle is to limit its size. Since evolution cannot be stopped, old services need to be removed.

This blog is moving very slowly (although one would think that there are enough provoking statements in this message to generate a few comments). I actually thought that it was dead, but there are a few readers coming everyday (that's what the stats tools say). I may accelerate the posting rhythm in a few months, when I'll be preparing a new course on the "Theory of Information System".

2 comments:

jeg said...

Your focus on theory of information caught my attention. Some of the problems you identified and are seeking to address ultimately arise from the use of poorly conceived methods of storing information, specifically those that separate data from its organization. A method that integrates them has the potential to store any kind of information and resolve many difficulties encountered when using the current methods. One method that integrates data and its organization is described in the "first principles" paper available www.varadata.com/papers.html

You have an aggressive research agenda. I wish you good luck with it.

John

Yves Caseau said...

Thanks for your comment and your pointer. I have printed your paper and will look into it right now. There are two difficulties with a "cleaner" approach, based of the "separation of concerns", such as the one you propose. First there is possibly a performance issue. I suppose this constraint will vanish eventually (maybe with your approach). Second it is usually difficult to implement when the information system is made of software packages that implement different schemes as far as data distribution is concerned.
Five years ago, I had a similar conversation with Serge Abiteboul, one of the world experts on database topics, and his advice was similar to yours. However, I thought that I could not put it into practice at Bouygues Telecom.
Cheers,
-- Yves

 
Technorati Profile