![]() |
|
Hands-on Software Architect |
Do you spend most of your time creating an up-front architecture and design with lots of artifacts(Architecture views, process and flow diagrams, models etc) and posting these architecture and design documents for reviews, updates, presentation to the management and development teams for feedback and comments? Or Do you like moving quickly to evaluation, validation and experimentation of your design and architecture by quickly implementing the skeleton architecture with the development teams and other stakeholders and following it with validated architecture refactoring and iterative evolution?Do you consider working skeleton-code with running tests as an architecture artifact? These are a few questions which define your architecting mindset and process followed to architect and design.
Software architects find immense satisfaction and pride in developing the architectural artifacts in isolation with the occasional ceremonial presentation. They find more comfort and security in their isolated ivory towers. Inspired by the real-world infrastructure architecture and architecting process, software architecture has been envisaged as a big up-front process, only for specialist – ‘An art’, which can be performed by fewer designated individuals. They communicate occasionally with artifacts created in isolation, and leave as soon as possible to enlighten other projects.
![]() |
Lonely Architect |
I see following reasons for this “Architecting-in-Isolation” mindset in architects:
- Software architects do not understand the boundaries of software architecture. They tend to do a lot of up-front design with limited feel for a ‘Just enough architecture’ which can enable quick design and implementation.
Undefined boundaries of Software Architecture - Software architects do not emphasise on continuous validation, evaluation and experimentation of architecture. Architecture does have elements which can be re-factored, iterated, and developed incrementally with more validated learning and feedback from stakeholders.
- Software architecture and design emerge and there are many unknown-unknowns. Architects tend to assume lots of things up-front. They cover it up with known architectural styles, patterns, past experiences and best-practices and ignore the fact that architecture and design will evolve and emerge as it get coded.
Unknowns of Software Architecture - The architecting cycle is very long, non iterative and incremental, and executed away from the development process. Architecting process is detached from the development process.
![]() |
Software Architecture and Development teams in isolation |
There are other reasons which keep architects away from the development teams. Like, involvement in other parallel projects. I will not discuss them here.
One big challenge architects always face is in deciding when to stop up-front architecture and design work, and start enabling design and implementation. Most of the time, architect tend to overload lots of detail in their early architecture artifacts. They attempt to define, every component, every interface, many views and system quality concerns; almost every minute detail that crop-up in their minds at the early phase of the product design. They tend to make lots of assumptions about the design and architecture without validation and verifications. Working away, in-isolation, from other stakeholders, they do not know what is just-enough optimal architecture and design for development teams to get going with implementation. They do not see the waste they are creating by big up-front design activities and impediments bought to the flow of agile and lean development of working and well-crafted software.
Architects needs to embrace lean and agile architecture and architecting process. They need to get their architecture artifacts consumed as soon as possible in the design and implementation iterations. They need to understand, that working code, which can validate their architecture decisions quickly and help them to evaluate design options is more effective than big architecture document created over weeks of isolated research. Now, I am not saying architecture should not be documented. Documentation can always be done incrementally and iteratively. Just lay down enough details, avoid long component diagrams, class details, pages,of flow and sequence diagrams, data models and deployment scenarios etc. Keep it simple, short, quickly communicable and consumable.
Even if it takes to write only just few draft pages, few wire-frames, hand sketches and board diagrams, Do it!. It has to be lean. Avoid all forms of waste which do not add value up-front. Design, in short iterations, communicate continuously, work closely with your developers and code for the skeleton architecture and cover it with the unit and acceptance tests. Take it to other stakeholders – product owners, Dev-ops, QA etc for reviews and feedback. Continuously re-factor and iterate over architecture and design. “Validated learning” about the architecture obtained through incremental architecture development and continuous validation is more effective than big up-front architecture and design activity which do not add much value and is very difficult to validate.
Now, experts will debate on how much you can actually re-factor software architecture. Well off-course, not to the extent that you can re-factor code, but yes, there are elements of architecture which needs to be iterated and re-factored. Like, components and service interfaces can be understood only after understanding the component usage and interaction patterns, service and resource relationships and dependencies. Interface design needs continuous refactoring. If you try to identify and design all your interfaces up-front, by making all architecture decisions at the start, just imagine the mess.
.
Functional models, Data models, thread models, development models, process and deployment models all evolve as you learn more about the system and its architecture and design. Sooner you get the working software, better you can evaluate and validate your architecture and design.
Which means architects need to be hands -on. They should be working closely with the development team to ensure that architecture and design is quickly and correctly translated into working code.
Try this to successfully achieve incremental and iterative architecture. Based on the known use-cases and user stories, list down some known high-level components, their relationships, important interfaces, some notes on desired threading models, sketches of data models, few non-functional concerns and constraints like failure scenarios, required performance with measurable numbers etc. Enable some early-communication and brainstorming discussions with the development team.
Even before you document details, sit down with the team to code the skeleton architecture components and elements with the acceptance tests for the required user stories. Define and name the system components and their interacting behaviour and interfaces. Define both the construction and interaction behaviour of the components. Functions are implemented with with no real code but well named classes and behaviours and covered with tests. Use ubiquitous domain language to define the components, classes and function names. Run the tests to ensure that component wiring is complete and correct with right interfaces as per the use cases. Ensure tests are developed and become the specifications of the desired behaviour.
The important thing for an architect is to know how much time she allocates for herself to get involved in day-to day implementation and design activities as early as possible and see the architecture getting codified. This is the only way to know how well the architecture is realized in design and code.
I believe, coding is a designing-activity and lots of design actually happens when we code. Can you see your architecture and design in code?
Try to answer the following questions for yourself:
- How much detail is enough to create and communicate architecture?
- Where do we think that the architecture is just enough for design and implementation to take over?
- Do you know about the risks associated with each element of architecture? Based on these risks how early you get these things designed and implemented? How much time you allocate to risky architecture component’s implementation?
- Do you continuously validate your architecture? And how?
- How much are you involved with the development team and how often do you receive the feedback on the architecture and design issues?
- Do you see architecture and design getting evolved as you code and implement? And how?
- What is your architecture documentation process? is it on-time up-front process?
- Are you still coding? Do you allocate time to get involved in implementation and coding? Coding exercise will help you to understand the architecture and design decisions better.