Skip to content
#Architecture

Modern Architecture 101 for New Engineers & Forgetful Experts - NDC Copenhagen 2025 (Category Architecture)

#Architecture #DistributedSystems #Management #Software #Leadership #Patterns #SystemDesign

Excellent. speech Jerry Nixon, product manager at SQL Server and Data API Builder, says architecture isn’t about “best practices,” it’s about contextual solutions and the courage to say no. In fact, Nixon breaks down myths and goes on to show you how to build architecture at any scale with your context in mind. He begins with a provocation, saying that the term “best practice” is used as an excuse for the absence of arguments. The only real best practice is not to use SQL injection. Everything else depends on: Company policies and budgets

  • Team level. (juniors vs experts) Legacy of systems and constraints from combining different systems
  • Possibilities to pay for mistakes (Twitter can do stupid things and you can’t.) Next, it shows the “annual rings” of technology: spaghetti code 90-x, three-tiered applications 2000-x and microservices 2010And then he concludes that it still works. As a result, architecture shaming is simply disrespectful of the context in which past architectural decisions were made.

According to Nixon, the role of the architect is to be the guardian of simplicity. After all, the architect is responsible for the most expensive decisions that can not be postponed. It's not someone who draws diagrams at the beginning of a project, but someone who constantly decides what to leave behind. The main task is not to add components, but to protect the system from excess - this is very similar to the thesis of Antoine de Saint-Exupery.

Perfection is achieved not when there is nothing to add, but when there is nothing to remove.

After that, the speaker begins to show how you can start designing a solution from the base and gradually increase the complexity as additional requirements appear.

  • Client → Database
  • Client → API → Database (add an interface)
  • Write API → Primary DB (synchronization) and Read API → Read replica. The logic is that you scale write and read independently by simply changing the connection string, and by eventual consistency. (It immediately becomes important for you to understand. consistency) Next comes the API Manager. (APIM) / Gateway for API management, versioning, load balancing, gradual rollout and a/b tests Then comes the monitoring story. (Nixon advises OpenTelemetry)cache and service bus Asynchrony is a way of working at scale: CQRS (Command Query Responsibility Segregation)Queue for working with load spikes
  • And so on.

So the talk is very interesting. The basic idea is something like All compromises come from reality, not ideology.

  • An architect is the one who gets blamed. Your decisions will be judged through 5 years when you have already left the company
  • The complexity has to be justified. Each added component (service bus, cache, gateway) This is another system that you are responsible for.

And finally the thesis from the author that I first amused, and then made me think

We're not writing code for a computer, but for the next developer, who might be an axe maniac. Don't make him angry.

#Architecture #DistributedSystems #Management #Software #Leadership #Patterns #SystemDesign