* RPC = function call
* API design = function signature
* data schema = simple structs
* message queue = list
* load balancer = simple facade object
* database = hash table / in memory SQLite
Once you have a rough sketch in place, you can run your prototype and test different behaviours.
It is a bit "tricky" to tackle asynchronous and concurrency, which you should do from the start. I prefer to do everything based on callbacks, and run my simulation from a static input list of events, ticking every component with a tick() method. This way I can generate different scenarios, save them to JSON, replay them.
Here's some examples:
- https://github.com/wesen/summer-pasture-netflix
cell.py is a kubernetes prototype (blackboard system), while app.py is the start of the netflix event notification system.
I am thinking of writing a course / book about the subject, and have a bunch of (wild) notes here: https://publish.obsidian.md/manuel/ZK/Structure+Notes/2+-+So...
I also played with pymc3 where you can just place plain python inside your probabilistic model and MCMC the heck out of it. It's slow but you can use some real world input and see what distributions arise at each point of the graph.
Often, decomposition is based on areas of domain or functionality. There are some downsides to doing it this way. The method in the book proposes it be based on areas of volatility instead. As an example, imagine writing a piece of software for FedEx shipping. What would change to use that same software for UPS instead? That may identify some areas of volatility. If you can plan that something might change, you can limit the ripple effect of those changes in future. Just an example of one benefit of doing it this way. The book explains some of how to approach designing this way.
Another point: Architecture is not just about designing the software. It is also designing the project to build that software. This involves things like critical paths, staffing, risk management, etc. Understanding these is important if you want to be able to deliver on time and on budget.
I suppose it is another tool in the box, however in the video it was about to be introduced as nothing less than a magic bullet.
Regarding the second point, I’m not sure I agree. An architect could get involved in staffing decisions and any project has a list of risks, for example I’ve maintained such lists as part of the architecture description. But I would define these these are side-tasks and not as core tasks.
Systems design is for me quintessentially embedded system architecture for a microcontroller, a SoC or combination. It could be a smartphone or console, or medical device, etc. This is a very broad domain… in my case I work exclusively on non safety-critical SW, so I’ve focused on learning Linux internals and system programming which I combine with an architecture framework. Typically I design middleware and daemons which work together to cover a specific area of a product. I work with different kinds of specialists (HW, kernel, performance, etc) and a lead SW architect for each specific platform. The system architect is a different person that oversees both SW and HW and requires skills in both.
For Linux I cand recommend the classics: Stevens, Kerrisk, Love. I found Chris Simmonds’ book to be a good general intro to embedded programming; it has much info which also helps with embedded system design. I’ve also picked up ideas from Making Embedded Systems by White or Embedded Systems Architecture by Noergaard.
Application design refers to individual applications running on an OS. These are the typical MVC or layered architecture styles. The official OS UI framework will likely define a recommended architecture while headless services are quite flexible and support several architectural styles.
Vol 1 of the Pattern Oriented Software architecture is a good intro. I have Software Architecture Patterns by Richards on my reading list, since it covers more recent styles like microservices.
You can spend years learning all about great design patterns that make sense for an IoT or embedded system but have completely backwards ideas about how to organize code for a game that needs to process 1000 unit movements per frame.
But once you figure out some narrower domain that you like, you should be able to find books about it , (if you're good at reading code) can start studying open source approaches to it, and can start prototyping small projects to gain some hands-on experience and muscle memory.
I don't know about the book, but I think MATLAB Tech Talk about systems engineering might be a good start for you: https://www.youtube.com/playlist?list=PLn8PRpmsu08owzDpgnQr7...
Part 3 is about designing a residential toaster :)
I've spent most of my career fixing bugs or improving existing apps/systems in C++. Every time I interview somewhere they want me to design a (usually offline) program or service and I don't know how best to do that. At this rate I'll be stuck at Senior Engineer for the rest of my career. (I'm already 14 years in).