Reference architecture, what is it?

I am at WICSA 2011 and the words “reference architecture” come up now and then in discussions. I am not sure I like the definition in Wikipedia, I think relies too much that you already have knowledge or experience of a reference architecture.
One simile (parable?) would be to a building code, a set or rules that underlie the actual architecture and construction of a house. The architect is free to build any type of house, as long as the code is adhered to.
But I like to use cooking as an example. A recipe is like an architecture, when you actually cooks the recipe and serve it is the implementation of software. You need to do certain things that are implementation specific, like tasting to see if the amount of salt suits your taste etc, and maybe you need to double all the ingredients to fit your  dinner party.
Julia Child writes in her book Mastering the Art of French Cooking that there are 6 principal ways to make a sauce, one being an emulsion of melted butter with egg yolks, which is common the common method for béarnaise, hollandaise and choron sauces. The type of emulsion is similar to a pattern while the three different recipes are designs.
For a restaurant; a product line architecture would be the menu they have based on the ingredients they stock, for example chicken could be used in more than one dish.
So what is a reference architecture for a restaurant? It would be the “rules” that guides and controls what is possible or desirable. It could be for example driven by the business domain, e.g. it should focus on French or Hunan cuisine. Woking as a “development method” would not be relevant in the former, but certainly in the latter.
Or it could be driven what is technically possible, e.g if the kitchen has no deep fryer it is impossible to for example make french fries.
Or there could be other restrictions, such a desire to only use locally produced ingredients.
Maybe I went to far with this simile (parable)?

Architectural views in practice

My research is not really focused on identifying the “best” architectural views for in-vehicle software. But as a side effect to the studies I am presently doing I got to think about three views that could actually make a difference in how well large projects succeeds (or fails). I have thought about the need of addiitonal views, but cannot think of an immediate need of more.

Integration is always an in issue in large projects that has some form of iterative development. With the increasing inter-dependencies between various parts of the system it is almost impossible to know what should be integrated when, for example what can actually be tested. There actually is already an architectural view defined that addresses this concern, the anatomy. I would describe the anatomy as visualisation of the complete system seen from an integration perspective, for example if a feature depends on a MOST interface it is no use to test it if the interface is not implemented. The visual picture means everybody should have the same understanding of the “.
The anatomy should dictate the order of development, delivery schedules and integration order. And the progress should be measured in how much of the anatomy is implemented, not in customer features. There is a relationship between customer features and the anatomy, but it is not one-to-one. you can read more about “Integration Centric Development and Anatomies” in a PowerPoint presentation or the article “Manifesting Shared Affordances in System Development – the System Anatomy” by L. Taxén and J. Lilliesköld.

The second view is the systems view, i.e. how the complete system is decomposed in smaller parts. I prefer to see this view as equivalent to the development view in Kruchten’s 4+1 views, i.e. it is aligned with the development teams. One can discuss if the system decomposition follows the organisation, as predicted by Conway’s law. The other extreme is that the development teams follow the most “logical” decomposition, i.e. ECUs, which is the most common unit to be outsourced to suppliers in the automotive domain.

The functional content of a vehicle is often defined by a list of customer features, which can be quite long, and this in itself can be considered an architectural view that concerns for example the business project, the marketing department and the end customer.
What is important for the development project is the relationship between this feature list and the two other views, i.e. what systems/development teams are responsible to implement the features and how the features relate to the anatomy. I don't know if these two relationships/mapping between views should be considered as separate views, which would make the total number of views 3+2.

What is a "good" architecture?

Hopefully you recognise it when you see it...

I believe every system has an architecture, documented or not, and intended or not. So I think it is reasonable to discuss if the architecture of a system is good or not, and that the question if the architecture description is good is a different, but related question.

So what criteria could you use to establish if an architecture is good? Is it suitable for it's purpose is a more precise question? But this probably differs depending on the stakeholder, a system could be very nice and intuitive for the user, but very difficult to develop and maintain for the developers. In the rest of this blog post I will writing about goodness from a developer perspective (including the architect, if one is involved).

I think there are different levels of how well the architecture suits it's purpose to guide and control the implementation.

The ideal level is where the developer understands the architecture and following the architecture is the obvious way to implement something. Doing it differently is seen as more difficult and less elegant.

The next level is where the developer understands the architecture, is able to follow it on her own and can evaluate if the implementation follows the architecture or not. The developer also understands the "debt" that would occur if she deviates from the architecture, for example in terms of compromising some quality attributes. At both this and the previous level the developer can contribute to the design of the architecture itself, and an appointed architect doing the design may not always be necessary.

The developer may need continuous support from the architect to be able to understand the architecture when working. This is very common and requires more work and puts limits on the the ratio between the number of architects and developers in large projects.

The worst level is where the developer cannot understand the architecture and the only one that can determine if the implementation follows the architecture is the architect by reviews. This is very work intensive on the part of the architect. The architect is also viewed as a police that interferes with the developers rather than supporting them.

I have worked at all this four levels in various projects, and I think that you need to be on the upper two if you talk about a good architecture from a developer perspective. I also think that if you want to be truly agile in a self-organising team it is at the two top levels you nee to be.

You tend to favour the solutions you are familliar with...

When I talk to students about the role of the architect I always make a point of the architect must know when not to use a particular solution/pattern/style (the old "Kill your darlings"). Regardless of this I have seen examples when an entire class thinks that a 3-tier architecture is the best solution in a project because they took a course on databases the previous semester.
Likewise, when I talked about component-based architectures in a lecture and gave examples based on AUTOSAR, a lot of students thought that components was the thing, even if it complicated the solution for the developers and the benefits with components was not really relevant in the student  project.

This is not surprising since I am talking about students with limited experience to various problems encountered when developing real systems. But I also think this is a problem for professional developers as well, including myself. We tend to stick to what we know and don't reflect if what we know actually makes things worse than better.

Is software engineering immature?

Is software engineering an "immature" engineering discipline? I have often heard this both in presentations and in reading, for example SEMAT states on the homepage:

"Software engineering is gravely hampered today by immature practices."

If I draw parallels between software engineering and civil engineering (arguably the most mature engineering discipline) my spontaneous conclusion was: Yes, software engineering seems to be more like how cathedrals were built in medieval times. The construction of them were based on rules of thumb and the practical skills and experience of architects and stone masons instead of the type of engineering practices taught at universities today. Interestingly enough, one can still see the "successful" cathedrals still standing several centuries later. But it would have been impossible for the architect of  Lund Cathedral to build something like the nearby Turning Torso. So the engineering in civil engineering has certainly "matured".
But where is software engineering failing? Everyone has heard of software projects running over time, over budget or have not been used as intended (or not used at all). But wait a minute! Exactly the same thing is common within civil engineering as well. There are many examples from a single contractor renovating a small house up to world known buildings such as the Sydney Opera House:

"The original cost estimate in 1957 was £3,500,000 ($7 million). The original completion date set by the government was 26 January 1963 (Australia Day).[16] Thus, the project was completed ten years late and over-budget by more than fourteen times."

So in this respect software engineering is as immature as other engineering disciplines.
So what is the fuss about? I agree that there is no consensus in the software community on how to do the equivalence of structural mechanics calculations or stress tests. Are we more hampered by this lack of consensus on how to do engineering tasks than by the difficulty to run large projects, which seem to be common to other engineering disciplines?

Recurring architectural decisions

A lot of academic (and industrial) research is looking at architectural decisions. One of the ideas pursued is that by capturing and managing this type of information it can be reused. I guess in practice it means that if you need to make a similar decision in the future you can look at what you decided the last time. I myself also saw this as an appealing idea. But today it struck me that we really don't have recurring decisions when we work as architects at Volvo. I think that if we had there would no be any need of an architect at all.

What we do have is a long succession of architectural decisions, each with a new set of prerequisites. The task as an architect is to be able to extrapolate from what has been done previously and still maintain some sort of conceptual integrity. To preserve the "feel" or "vision" of the architecture while still fulfilling necessary functional and quality requirements (including cost) and operate within the constraints given by for example our organisation and what is available/possible form suppliers.

Architectural modelling

To have a more formal definition of what I mean when talking about modelling an architecture (summarised from to IEEE 1471):

An architectural description is organized into one or more views.
A view may consist of one or more architectural models. Each such architectural model is developed using the methods established by its associated architectural viewpoint. An architectural model may participate in more than one view.
The viewpoint determines the languages (including notations, model, or product types) to be used to describe the view, and any associated modelling methods or analysis techniques to be applied to these representations of the view.

I am not against using models to represent architectural views, but there are some fallacies that often seem to happen when using various modelling approaches in architecture descriptions.

• The purpose of the architecture model is unclear (in 1471-terminology: It is not obvious what concerns the model addresses). UML is especially problematic since it provides notation that can be used for so many purposes. Read John Daniels paper Modeling with a Sense of Purpose as an introduction.
• Everything is modelled to the same level of detail. Not everything is equally important to know. For some parts of the system it is important for the architecture (and realisation of qualities) to know the details, for other parts it is not.
• Every part of the system must be represented in the architecture, for example if every class must be traceable from some element in the architecture. This could even be a rule of the architecture, "if it is not explicitly allowed it is forbidden". This reeks very much of waterfall mindset with the architects doing much of the design before any developers get involved. I think it is presumptuous  of the architects to claim to know everything that needs to be implemented.
• Modelling tools always demands a precision in notation, which is not always desirable when sketching an architecture. You cannot model "sort of a class", or "could be a function" when using a tool, which is possible on a whiteboard or a napkin. In a tool it has to be a class, a function or a package, not something in-between.

Architecture for small systems?

It is Friday afternoon and I'm allowed to have some wild ideas...

After a long e-mail conversation with one student of appropriate architectural styles to use in a project course I realised that his ending comment is spot on:

I think the course and it's literature are focused on large systems and seeing that we are new to this, especially with a new language in parallel, its difficult to not apply what we learn during the course.

In my opinion the patterns covered in the book left me half way with the idea that most architectures could fit into a pattern, within its own right.

I have been too blind when I teach, and not even my professional experience helped me to identify the problem:

Most (all) literature about architecture teaches solutions for big systems.

Is there a niche for information on how to architect small systems? Small in the sense of not having millions lines of code, or a large development team, or a long project. I know that the agile manifesto states that that one of twelve important principles are

The best architectures, requirements, and designs emerge from self-organizing teams.

But do a google search on "agile architecture emerge" and see what comes up, a lot of interesting reading that suggest that the issue is not that simple.

A lot of software systems are small, from a bittorrent client (the student project) to the software in the door control unit in a car, to a mobile app. And yet they would all benefit of having some though about what they must meet for non-functional needs that is addressed by a (simple) architecture. Should I write a book on architecting small systems? So that a team member is better prepared when he or she participates in a small project where the architecture "emerges"? You don't need to be an architect to benefit from doing thinking at an architectural "level".

Quality attribute scenarios

Defining a quality attribute, a.k.a. non-functional requirement, is not easy. I know since i did that for a new platform at Volvo. When the verification and validation group comes back and says that it is not possible to verify this as a requirement I can only assume that I did not do a good enough job. But I have to agree with what George Fairbanks wrote in his blog:

Before you get too excited, you should know it’s easier to write these for quantitatively measurable qualities (e.g., throughput, latency) and harder for softer qualities (e.g., modifiability, usability).

There are two recent books that have a more agile approach to architecting: Lean Software Architecture by Coplien and Bjørnvig, and George Fairbanks' Just Enough Software Architecture: A Risk-Driven Approach. Unfortunately the latter isn't available in any Swedish internet bookstore, but as soon as I get hold of them I will post a review, as I have reviewed other books on software architecture.

How attractive it is to work as a software architect...

A colleague sent me this link from CNN Money about how attractive it is to work as a software architect. Go figure...

Hidden leaves

Philippe Kruchten wrote a very entertaining interpretation of Tao Te Ching from the perspective of an software architect. But besides being entertaining it has some very good insights into the nature of an architect, like the following:

The architect doesn't talk, she acts.
When this is done,
the team says, "Amazing:
we did it, all by ourselves!"

Personally I'm more familiar with Japanese texts (more than 20 years of budo training) so I thought of interpreting the Hagakure with the same perspective (you can find a translation to English of the original text here).

Among the maxims on the architect's wall there was this one: "Matters of great concern should be treated lightly." A senior developer commented "Matters of small concern should be treated seriously." In one's project there should not be more than two or three matters of what one could call great concern. If these are deliberated upon during ordinary times, they can be understood. Thinking about things previously and then handling them lightly when the time comes is what this is all about. To face a problem and solve it lightly is difficult if you are not prepared beforehand , and there will always be uncertainty in hitting your mark. However, if the foundation is laid previously, you can think of the saying, "Matters of great concern should be treated lightly," as your own basis for action. (1st chapter, p. 33)

According to a senior developer, even a poor programmer will become substantial in the art of writing code if he studies by imitating a good model and puts forth effort. An architect should become substantial too, if he takes a good architect as his model. (1st chapter, p. 40)

The proper manner of architecting is nothing other than not being careless, but in this way one's design will be sluggish and stiff. One should go beyond this and depart from the norm. This principle applies to all things. (1st chapter, p. 48).

In carefully scrutinising the projects of the past, we find that there are many different opinions about them, and that there are some things that are quite unclear. It is better to regard such things as unknowable. An architect once said, "As for the things we don't understand, there are ways of understanding them. Furthermore, there are some things we understand just naturally, and again some that we can't understand no matter how hard we try. This is interesting."
This is very profound. It is natural that one cannot understand deep and hidden things. Those things that are easily understood are rather shallow. (1st chapter, p. 69)

Software quality attributes

I got the following e-mail from a student after discussing definitions and frameworks of quality attributes:

Hello,

It seems there are 2 main groups of ISO standards dealing with software quality:

1. a 4 part ISO 9126 standard (QA considered are reliability, usability, efficiency, maintainability, portability AND, actually, functionality)
2. second generation of ISO software quality standards referred to as SQuaRE (Software product Quality Requirements and Evaluation): ISO/IEC 25000:2005, ISO/IEC TR 25021:2007 and ISO/IEC TR 25060:2010.

I couldn't find any free copies of these standards, so we can't really use them.

Anyway, it would be interesting to get during the lectures some additional information about different quality attributes models out there (early McCall and Boehm models, FURPS, ISO models etc.).

I replied with this:

You are correct. You can't access the ISO standards (without paying).

• For a very soft introduction to quality attributes, read this: http://www.stellman-greene.com/2009/10/03/using-nonfunctional-requirements/
• An in-depth analysis of a few quality attributes are available at http://www.sei.cmu.edu/library/abstracts/reports/95tr021.cfm
• For a definition of terminology I suggest IEEE std 610.12, bit this covers much(!) more than just quality attributes. It is available at http://ieeexplore.ieee.org/
• You can also read what the PhD students at Blekinge Technical University wrote down in a course about software quality: http://www.bth.se/tek/besq.nsf/pages/017bd879b7c9165dc12570680047aae2!OpenDocument
• And as usual, wikipedia is a good introduction, look at "software quality" or "FURPS" for example.

I think it will be a little too much to go into detail about different quality frameworks in this course, above they dedicated an entire course at PhD level to the subject.

Hope this helps,

Ulrik

Why I don't like formal methods...

I know the title of this port is a bit controversial and with it I alienate a lot of researchers, even at my own department. But I will try to clarify my arguments and hope that someone can prove them to be wrong.

• To be more precise; it is not the formal methods in itself I dislike, but the prominence they seem to have at prestigious software engineering conferences. I think it is not at all in proportion to how well-used formal methods are among professional development.
• Formal methods are really attractive from a researchers perspective. You can use concepts as theorem, proof and deduction and other nice things. But nice is not the same as relevant.
• Even though I have heard about formal methods as one of the enablers to establish software engineering as a "real" engineering discipline for almost ten years I still don't think they are nearer being well-established now than then.
• Some claim that you can never be sure of the software unless you can prove properties about it, and I agree that presently formal methods are the only technique that promises to do so. But there is a lot of successful software out there which have not been proven.
• Specifications that fulfil the requirement of being interpreted formally are hard to write. Compare it to learn a new programming language and be proficient enough to use it without any side effects.
• I don't know if formal methods scale well. It is one thing to use it on a nice prototype system with 30 entities developed in your lab. It is another thing to use it on a system with 500 entities, many of these with quite varying quality in requirements and code.
• I still don't know of any that uses formal methods for real; i.e. as part of the normal way in products shipped to customers in business intent on making money. Not in one-shot attempts, pilot projects or by non-profit organisations.
  A search on google skolar of industrial+software+formal+methods list papers from the 90s as the top results. And these seem more to be arguments against what I claim above rather than actual reports of usage. I really would welcome examples.

Tacit knowledge

At the ECSA conference there was a lot of talk about tacit knowledge and the importance of tacit knowledge when architecting systems.
I completely agree that one thing that differs an experienced (and productive) architect from a fresh graduate is is the amount of tacit knowledge the former possess. However I thought that in the discussions at the conference there was some confusion of what is exactly meant by tacit knowledge.
On one hand there was the view that tacit knowledge was simply the architectural knowledge that was not documented, i.e. stuff that only resided in the peoples' heads. To improve the management of this is mostly a question of capturing it in the right form and with the best tools.
On the other hand one sees tacit knowledge is such knowledge which is difficult to express using language, which is the tradition I'm used to.
In order to capture the latter type of tacit knowledge it is not just a question of having the time to do so or the right tools. The difficult part is for the person having this knowledge to be able to formulate in in such a manner that he can express it.

Concurrent architecture

To continue my thoughts about the programming language Erlang and especially how an architecture for a system implemented in Erlang would look like...

Joe Armstrong's PhD Thesis provides a very good understanding of the basic principles of Erlang. I think that some of the mechanism in the language makes several pattens in Pattern-Oriented Software Architecture Volume 2: Patterns for Concurrent and Networked Objects unnecessary.

I think it would be quite useful to describe the architecture in views. I think the "standard" 4+1 views could be used quite easy for a system implemented in Erlang as well.
However it may not be the logical view that the other views are derived from, rather I would think that the process view would be the basic view. Kruchten has an underlying assumption that the logical view is object-oriented. This is obviously not relevant for a system implemented in Erlang so a different notion of the logical view is necessary.

Joe Armstrong strongly suggest a hierarchy of processes. A problem domain model could be a good way to identify the top-level processes, e.g. the PDM classes could be the initial processes and the asynchronous interaction will be implemented as messages, since reality is asynchronous. I need to try this in practice though...

More books on architecture

I previously wrote about some textbooks about software architecture. I have have since received three more books.

The first, Software Architecture: Foundations, Theory, and Practice, I disregarded fairly quickly. I think it is too verbose (712 pages!). I could not find the brevity and clarity I think a good textbook should have to should support understanding instead of overwhelming the reader with facts.

I have not had the time to read through The Art of Software Architecture: Design Methods and Techniques yet, it arrived today in the mail.

Software Architecture and Design Illuminated covers a number of different types or styles of architectures, and does so in a manner familiar to the students at SE & M (i.e. in UML models and in Java code examples). It takes a completely different approach to quality attributes compared to Software Architecture in Practice (SAiP), it is structured according to different architectural designs and explains what they are good and bad at. So rather by learning by reason about concepts if focuses on learning by examples. This means that the coverage of quality attributes and tactics is not as comprehensive as in SAiP, for example attributes and tactics related to embedded systems are lacking in comparison since there are few examples of those in the book.
However since the course presently is structured according to theory and not examples of different types of architectures it would require more update of the course compared to the other texts.
The book also has some on-line material for instructors which I have not yet been evaluated.
Summary: Good book which I think presently is the best undergraduate book out there. I like the example-based approach, but it might just be appealing to me because of the fresh approach to an already familiar subject.

On variants...

During my research I gained some inside information into how various automotive manufacturers work. One very interesting observation is to see how different companies approach variants.

I know of two high-end companies having as a general architectural strategy not to have variants of systems or ECUs. Note that the electrical system as a whole may me different due to e.g. feature content, but if the system has a body control module there only exist one variant of that module both in development and manufacturing. To my understanding they are only using the optional variant in F. Bachmann and L. Bass, "Managing variability in software architectures," SIGSOFT Software Engineering Notes, vol. 26, 2001, pp. 126-132, and not any of the other types.

Personally I think that this is very beneficial for a company, both from an engineering and from a manufacturing perspective, even if it seems wasteful not to cost-optimise an ECU.

But this strategy is probably not for everyone. There are some common characteristics between these two companies:

• The are both high-end brands and have customers that can afford to buy the best.
• They are not mass producers, only in the order of 100k vehicles per year. This means that development costs are not negligible to article cost, even if they are smaller.
• Tailoring to the customer needs is an important business strategy.
• The have unusually high ROI compared to the average for the automotive industry. Being successful is another way of describing it.

And no, Volvo Cars is not one of the two companies...

The question is how you "dare" to go for such an architectural strategy if you are not already doing it? I think developing a business case to show that you would earn more $$$ with such a strategy would be extremely difficult.

Fallacies towards lean architecting

If architecting is seen as an effort on which you can spend more or less resources with a better or worse result I wonder how you identify the trade-off where you get the "most bang for the buck"?

I think this is a fundamental question to ask if one discusses architecture from a lean perspective. To use the standard lean criteria, to eliminate everything that does not contribute to end-customer value, disregards several important (?) quality attributes of an software architecture. One could argue that a product line approach benefits the end-customer, but to be honest I think it is the developing (creating?) organisation that reaps the main part of any benefits.

What are some fallacies that would detract from the most efficient architecture? A personal list would be something like this:

• Completeness of the architecture - To aim for an architecture that describes 100% of the system at some abstraction level is not efficient. I don't think I need to detail this argument besides Pareto's law.


• Formality - Strictness in describing the views may hinder the stakeholders' understanding of their concerns. To much effort on defining and/or understanding modelling rules is not efficient.


• Detail - An architecture is a top level design, and I would put an emphasis here on top-level. It should not bother with details best left to other tasks (and stakeholders)


• Traceability - To require traceability to all requirements that affect the design of an architecture and traceability from all requirements resulting from architectural decisions is not efficient. On the other hand to have no traceability is not efficient either.
  There should be enough to convince the stakeholders that vital concerns have been addressed. So it basically is the stakeholders that define the necessary level. Note that very rigorous stakeholder may completely big down an architect with demands on traceability...

I would like to expand these thoughts in a more rigorous way, enough to present at some workshop, or at least have more discussions with researchers and practitioners. I'm still waiting for the book by Coplien and Bjørnvig.

Some more thoughts on agility and architecture

This is a quote from Grady Booch's blog:

All architecture is design but not all design is architecture. Architecture represents the significant design decisions that shape a system, where significant is measured by cost of change.

One interpretation I can make of this quote is that it seems to be an inherent contradiction in aiming for agility in the design decisions that constitutes the architecture. Architecture is those decision that you don't want to change since it is costly to do so (in $$$, man hours or lead-time). But this is not the same as making those decisions up-front.
Maybe a lean approach is what benefits agile development the most:

...an architect should make as few decisions as possible, deferring the rest until later in the lifecycle.

Tyree and Akermann in IEEE Software 22(2), 2005, pp. 19-27.

SEI has a webinar 22 April about Agile Development & Software Architecture. Maybe that can shed some more lights on how to reconcile architecture and agility. To bad they are broadcasting it at such an awkward time for us in Europe.

Is software a product of creativity? Or?

I think the phrase "development process", or rather the concept of development as a process has become a paradigm in software. A google search on software development process results in 724 000 000 hits!

The view of creating software is done through a "development process" is so established, at least in the software industry, that no-one seem to question if this is a valid view on how to design stuff. This regardless if one talks about waterfall, spiral, agile or lean.

Since I am not a native English speaker it could be a lack understanding the connotation of the words, but I'll try to explain what I mean.

First, software is developed. No-one talks about a software creation process. To me development is more akin to refinement, exploitation or progress. I think software is created, it is one of the very few intellectual pursuits besides the arts where something is created from nothing.

Second, the development (or creation) is done through a process, something which in an engineering perspective means that by combining the right inputs and having the right circumstances, a desired output is achieved. And the process is repeatable as long as the inputs and circumstances are the same. I don't think anyone disagrees when I say that not two software projects are alike, regardless how similar one tries to keep the process.

So the question is how should we view software creation? I think it would be helpful to see it as a creative endeavour similar to what a writer does. There is usually some idea of what the end result should be, but that may or may not be very detailed when starting to make sentences. Some write in a very linear fashion, starting with the first chapter. Some start with an outline which gets more and more refined until it actually is the full text. Some writers are very disciplined and can write for 8 hours day, some can just write when they are inspired. And everybody has heard about "writer's block"...

If software is not developed in a process, but seen as a work of creativity I think efforts should be to make the people writing software as creative as possible (and therefore also productive). Maybe this can be used as a "lean" principle on how to manage a software project.