SDD Conference Top Takeaways - Evolutionary Architecture
- SDD Conference Top Takeaways - Evolutionary Architecture
These are just a few of my takeaways from Neal Ford’s Building Evolutionary Architectures workshop at SDD Conference 2022. More to come.
- Software degrades over time, like cracks appearing in a building. Structural defects can start small and get more problematic. This seems inevitable, but it isn’t.
- QA helps to reduce degradation, but it doesn’t cover everything. It misses the architectural characteristics, e.g. auditability, scalability, performance, etc.
- A full list is not possible, they change over time. You need to create a list specific to your organisation and iterate on it.
- Ensure you use ubiquitous language within a project team and across the architects in an organisation.
- Tech change is pretty much forced upon us.
- If you don’t evolve or pivot you get left behind.
- Patterns emerge and are given a name when they are recognised as a good solution to a problem.
Q: How is long-term planning possible?
A: Plan for things to change!
“An evolutionary architecture supports guided, incremental change across multiple dimensions.” i.e. architectural characteristics that are relevant to the project.
- Automating checks for architectural characteristic support work in the same way as unit tests.
- Higher level functions will require possibly stringing things together, as there isn’t a specific library or tool that can be used.
- You are creating an objective integrity assessment of one or more architectural characteristics.
- Iterate. You don’t know what you don’t know. Things will change.
All architectures become iterative because of unknown unknowns; agile just recognizes this and does it sooner. - Mark Richards
- Fitness functions overlap with other premises, e.g. monitors, unit tests, chaos engineering…
Some Examples of Architectural Fitness Functions
- Cyclic Dependency Function - ensure you don’t have e.g. ClassA referencing ClassB that references ClassC, which references Class A… This is a small thing that can be missed and leads to Bit Rot.
- Directionality of imports - ensure rules on this are followed.
- Data displayed on screen is not stale - ensure HIPPA is followed for US healthcare systems.
- Chaos Monkey & the Simian Army - check your system copes with systems, availability zones and even whole cloud regions being taken down.
- Naming conventions - ensure that naming conventions are followed by specifying a pattern to use or antipattern to avoid.
- Afferent (incoming) and efferent (outgoing) coupling - check whether coupling falls within specified limits.
- Identify changes in open-source software licenses - make sure any changes in open-source software you depend on are put under the noses of your legal team.
Developer Torture is not the aim
- Architects and developers need to work together.
- Fitness functions are a checklist to stop important things falling through the cracks. Doctors use checklists too!
What makes microservices so evolvable?
- Extremely lose coupling.
- Fine granularity / small quanta, including any data stores.
Implementing Fitness Functions
- Identify your dimensions (including architectural characteristics and linked metrics)
- Define fitness functions for them.
- Use deployment pipelines to run your fitness functions (quick and atomic functions running earlier than slow and holistic functions)
- Beware of the Last 10% antipattern! 80% of requirements are implemented easily, the next 10% are hard as the tool/platform doesn’t really support them, the last 10% are impossible. Low/No Code is prone to this, otherwise we would all be using it!
- Try the hardest thing as a proof-of-concept (PoC), not the simplest (we would hope it can do that easily!). Will it really do what you need?
- If it acts like code it should be handled like code with unit tests, versioning, etc.
If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.
- We’re told low/no code isn’t for developers so it doesn’t integrate with pipelines, version control, etc.
Some Tools for fitness functions
- Jupyter notebooks: document and have fitness functions together. Executes on opening.
- Architecture Decision Records (ADRs)
- Building Evolutionary Architectures by Neal Ford, Rebecca Parsons & Patrick Kua
- Refactoring Databases by Scott W. Ambler & Pramod J. Sadalage
Let me know if there is anything specific you would like me to expand on. More to come from SDD 2022.