Sam Newman on Architectural Simplicity
In a recent talk on software design, distributed systems expert Sam Newman argued that ambiguity in ownership is the biggest bottleneck as systems grow. He stated, “Simplicity isn’t just a principle—at scale, it’s an architecture feature.” The remarks emphasize the need for clear domain boundaries and standardized interfaces to manage complexity in large-scale backend infrastructure.
Sam Newman advocates for microservices only when the complexity of a distributed system is justified to solve a specific problem, such as teams getting in each other's way. He often surprises followers by recommending a "monolith-first" approach, as monolithic applications are not inherently flawed and are simpler to work with initially. The clear domain boundaries Newman mentions are rooted in Domain-Driven Design (DDD), specifically the concept of a "bounded context." This approach models services around distinct business capabilities—like payments or inventory—rather than technical layers, which is considered the most critical and challenging design decision. Clear ownership, where a single team is responsible for one or more services, is a core tenet for reducing operational friction. This clarity directly leads to faster incident response and resolution (MTTR) and incentivizes teams to proactively manage technical debt. A key principle for maintaining these distinct boundaries is that each microservice must own its own data. Sharing databases between services is an anti-pattern that leads to tight coupling, and architects should avoid designs that would require complex distributed transactions. Newman draws an analogy of a software architect as a "town planner," who should focus less on the internals of each service and more on how they are connected. The primary concern for the architect becomes the design of APIs and the integration patterns between the "zones" of the application. For migrating from an existing monolith, he champions an evolutionary approach rather than a complete rewrite. One well-known strategy he details is the "Strangler Fig" pattern, where new microservices are built around the old system, gradually replacing its functionality over time.
