The place of architects in a business organization
[Author : Kinshuk Adhikary - software plumber.
LinkedIn : http://in.linkedin.com/in/kinshuka
Blog : http://me-plumber.blogspot.com/
Email: email@example.com ]
Introduction : Due to a profusion of roles and titles in the software world, there seems
to be some confusion about the architect, and the place and roles of such people in
This write-up is mainly for business people who may appreciate a perspective that is a
mix of the technical and the management aspects of this problem.
Disclaimer : The reader must clearly understand that everything written below could be
inapplicable, that everyone's experience is different, and a lot depends on the type and
nature of the system and the business. My own experience has been in business
products and applications, development, launch, adoption and modification cycles. You
will need to form your own opinion on the correctness of everything said below.
Understanding architecture : The simplest definition that I can come up with, cobbling
several smart people's views and misquoting them, is - "Architecture is an abstract
description of the system in its various working phases and in the various stages of its
In the largest sense, the "system" is the entire organization, seen as an information
machine. Quite naturally, the architectural responsibility becomes not individual small
systems, but policies, standards and quality aspects that apply to every system in the
The most important external responsibility of the architect is communication. Without
clear communication, both to technical and non-technical teams, the inward
responsibilities i.e. the abstraction and the description , lose their meaning.
An architect needs to constantly distance oneself from the actual "concrete"
implementation of the system. Too much contact with concreteness disturbs the
abstraction that is extremely needed.
At the same time, at a private level not always talked about, the architect needs to be
extremely hands-on in certain "pilot" areas. Too much distance from actual
implementation makes the architect quite unable to reach out to the real implementers -
who are (a) the design folks (b) the coding folks.
Another disclaimer : As I said earlier, the term "architect" has been extremely misused
and misapplied, especially in the last 5-6 years. While writing this, the architects I have
in my mind are those I have met and worked with "more that six years ago".
Phrases like "an architect should not do any actual implementation" can get extremely
distorted if your picture of the architect is not the same as mine.
The kind of architects I am talking about were (a) very highly technical, up to date and
extremely knowledgeable in all aspects of their work (b) very busy all the time, far more
"work" than they could handle (c) visionaries, almost mystical - I do not know how to
express this. Very often, the systems they created were nearly perfect, and brought to
the organization and stakeholders extreme successes and large amounts of money.
The committee, the council, the guild : It is extremely unbalanced to have a single
architect in charge of a large mission-critical system.
Because of high costs, organizations often think a single architect is all that is needed.
But a single architect could be a disaster for the system or the product.
Everyone in the organization must get used to talking in terms of "the council of
architects", a group, a committee. All actions must be attributed to this group. There
usually is a Chief Architect, a key spokesman who enjoys the confidence of top
managements. But architectural decisions are taken by "the group", during internal
meetings, in consultation.
Why is it unbalanced to have a lone architect ? Because of too much uncertainty, too
vast a field, the need to discuss and thresh out between peers. And a hundred other
reasons, I will write some of them below.
- Profile of the individual : Most architects I have known have been a mixture of immense
egos and extreme humility. It is part of the training and the drive that is needed to keep
up to date and on top of everyone else. In a group of peers, a moderating effect sets in,
and many ego-driven unidirectional mistakes are avoided. As every architect is aware,
the number of things one does not know exceeds the known by factors of 10s, or 100s.
- Peer understanding : This has a spurring effect on an architect's abilities, it is all about
bouncing things with others who resonate on the same frequency, are of a similar
background, although of slightly different specializations.
- Policy and guidelines : Beyond the immediate system, the work is a lot about policy.
One cannot set a policy without a very holistic understanding, almost impossible for one
person to do.
When does an architect work in a silo ? : It is also possible that an organization has
many architects each working in a silo. This is typically the case when the org has many
business units more or less independent, and each catering to a different class of
customers and user groups and lines of business.
However, sooner or later, as the "enterprise" view of the organization gains priority in the
minds of top management, a council does get formed, all systems in the organization
follow some core principles, initially very very few.
The lesser the better : An architect's work obeys this more often than not. Too much
detailing is bad. Too descriptive is bad. It is more important to "hide" the complexity
behind abstract black boxes.
(purely my own arbit statement) - Good architecture seems to be very cognizant of
Godel's incompleteness. For a system to be both complete and consistent over time, its
complexity has to be achieved rather slowly, the relations between its parts and the parts
themselves cannot be written in stone too quickly.
At the same time, the complexity is all there, for whoever wants or can dig deep enough.
It exists as a mass of "maybes", not as "musts".
The best things are incomplete "shells", into which the other participants, the design
folks and the developers, can put in their best notions.
Important to appreciate - 3-year paradigm shifts, 6-year eras : Good architectural
styles (the decision driving large system efforts) are those that can reasonably withstand
the flux of technology.
Technology changes very fast (ugh!!) - a paradigm change happens every 3 years, and
an era changes every 6 years (era == my own term, two paradigm shifts, if you are not in
touch, you are in oblivion and it is too difficult to catch up).
Lets say that in 2000, JDBC was everyone's staple. In 3 years ORM changed how
everyone looks at data persistence. Now it is the age of Hadoop in the cloud. So which
is your particular cave ?
The problem is not that old things will not work, but that applications, systems built using
them will get slowly isolated. When several eras have gone by, the great old stuff will
now become heavy weights making it unable to do the nimble things that everyone
around will be doing.
That is another reason architect's are both enthusiastic as well as careful of technology.
And why individual architect's can be too prone to "push" only those things they have
had personal experience in, and why a council becomes necessary.
It is of course needless to state that an evolutionary progression of knowledge, right up
to the here-and-now, is a necessary toolkit for every architect, excluding nothing by
personal preferences, including as much as is humanly possible.
[P.S : An architect of my acquaintance had to take a 15 day vacation, just to study
ORM/JPA/EJB3 in which he was feeling out of depth. He was about 50 years old, and
was cursing the day he took up this profession, and was warning other young techie
enthusiasts who all wanted to be architects].
Sir Christopher Wren : Was a good mathematician, had extremely good mechanical
skills, was a good water color painter (or is it sketches, I don't remember, look up
Wikipedia), and enjoyed the confidence of the patrons who wrote the cheques. All in all,
an ideal architectural scenario.
So who is this "design" guy ? : Enter the next participant in the process of creation.
This is the person the architect finds it easiest to talk to (other than other architects, of
course). They bridge the gap between the abstract and the lab experiment.
The design person is a terrible realist, who deals with artifacts of all sorts, UML and what
not, and the latest APIs. The whole focus is on solving a problem in the best available
way within the description the architect has provided. Sometimes both are the same guy.
Architects are often wrong. That is obvious, if you are not about to make mistakes - you
are probably doing nothing of any value to anyone. For example, EJB1 was most
probably a mistake. But EJB3 is most likely not a mistake. No one really knows 100%.
(Side note : Personally, I get very worried when a Project Manager says - oh, the project
will take 6 months and 17 days, and there are only six risks, and here is the mitigation.
Does it really work that way ? Does this guy know anything at all ? These definitive
numbers are the warning signs of complete ignorance about complexity. If the project
does get done in 6 months and 17 days, then you can be quite sure it is not the same
project that was initially discussed, but a much diluted, much changed version, whose
internals only the developers know which they guided as they wished :-)
It is the design people who go a little more concrete, do more pilots and initial studies,
and more or less assure that things are most likely going to work out when the
developers write the final code.
More about design folks and developers in a later discussion. It is sufficient to say here
that design and development are in their own rights equally important to the effort. The
ridiculous hierarchies that emerge in organizations (developers are way below,
designers are somehow "above", and architects are the "cream de la creme" - these
things one wishes did not exist).
For the business : Why is an architectural council even needed ?
The basic need is "direction". There is a huge technology flux out there, you need some
ways to withstand it. Or the best point of entry into it, given the stage of evolution the
organization is in. Do you really need SOA ? Why should you go into the cloud ?
Cutting across the jungle of jargon, you need to evaluate the suitability of various
alternatives that best suits the business as it is now and as it will be in 3 years and 6
years. If you are in the client-server era, what is the best way to jump in ?
[I first learned about the cloud nearly 3 years ago, from an architect I respected a lot. He
himself did not know enough about it back then, I guess, but did his best to explain.]
Of course, solving the immediate problems, conceptualizing a solution, overcoming the
usual architectural constraints of performance, scalability, extensibility, achieving
standards and quality, consolidating across systems - these are other (and sometimes
minor) needs that architects fulfill.
Most applicable when the organization is seen as an information machine, whose
competitiveness depends to quite an extent on the systems that process information. But
then today, such a view is applicable to nearly all organizations. There are still many
who believe that systems purchased from vendors without an internal effort will solve all
problems, but the jury is still out on this one. Most are embracing data analytics, or soon
"The architect" was interestingly described in Matrix-3, but of course business needs
some help, maybe like this write-up, to distinguish between a movie and the real world.
Otherwise it is easy to fall into the trap that the concrete is the real. It is not so. It is the
math that is the real. Ever heard of "high frequency stock trading" ?