Dealing with Marchitecture

cartoon by david fletcher - cloudtweaks

marchitecture (aka marketecture) is an architecture produced by a vendor mainly from a marketing perspective. as a company's engineer, architect or manager, you will be exposed to it at some point or another. vendor representatives will come in and make a shiny powerpoint presentation about a new product, and you and other colleagues might be asked to provide your opinion on whether that product could be of any use in your company.

i used the word "shiny", because the main point of a marchitecture is to get your attention, so the vendor can engage in further conversations & negotiations, and...make a sale, of course. here's an approach that can be used to deal objectively with a marchitectured product.

first, start by not making assumptions .

do not assume, simply because it's a marketing presentation of the product, that it's all a bunch of crock punctuated by catchy buzzwords to make up for the smell . at the opposite end of the spectrum, do not assume that the polished powerpoint view reflects reality.

one consistent thing marchitecture does is emphasize the positive aspects of the product, and at the same time "forget" to mention, or downplay the negative aspects.  vendors do not usually lie, not because of particularly high moral standards, but because it is bad for business in the long run. they don't lie, but they don't tell you everything you should know. with that in mind, evaluating a product can be a challenge. there are however a number of things you can do during and after the presentation:

1) during the presentation

first, the very basics. boxes with pretty colors do not make an architecture.  look closely at  the architectural layers in the presentation diagrams. are they clearly separated and obvious?  focus on the flows drawn between those boxes,  look for cyclic dependencies , i.e. if  the lower levels are dependent on the upper layers. here are some more elements you may want to consider in your evaluation:

• pay special attention to diagram components where the vendor talks the least about. might be the weakest link(s) of the product.
• consider the time factor: how will the system evolve? does it take into account the development life cycle ?
• does the product clearly and completely address the business problem it is supposed to solve? if not, what is missing from the diagrams ?..
• what are the product dependencies ? what external components  (e.g. by the same vendor) does it need , to function properly?
• etc..

2) after the presentation

ask the following questions in order:

• does this product address your specific business problem(s) ? it might actually solve issues you do not have, or solve issues you have but only partially.
• if so, is it a finished product , or at least are some parts of it available for you to test ?  some products are widely marketed but...do not exist yet. we need working software, not slideware or vaporware .
• if so, is the product or a subset of it available for testing within your company's own settings, and for a sufficient period to assess its worthiness?
• are there some additional & more detailed documentation geared towards more technical audiences (eventually free of marketing spin)?
• while you're at it, ask for references . who else is using that product? is there a forum where eventual issues with the product are reported & discussed?

as for any serious software evaluation, y ou need to do a trial of the product  in your own enterprise environment before eventually buying it, no matter how great it  looks on paper, and how famous & well-established the vendor, and how outstanding your relationship with them. it really is the only way to objectively assess the product's adequacy for your problem domain . even if it does in fact address your particular needs, maybe the product will create more issues ( accidental complexity ) that it will solve ( essential complexity ), and that would be grounds for rejection.

the matter becomes a bit more complicated when you have one or more technology bigots (a variation of the corncob anti-pattern ) at upper hierarchical levels, who do not particularly care about points of view other than their own. in some cases, they might have already taken the decision to adopt or reject the product before the presentation even began...and give you and your colleagues the illusion of participating in technical decisions. regardless, your job is to do an objective evaluation . whether that evaluation will be taken into account is another matter.

source: http://tsicilian.wordpress.com/