A practical discussion on when it actually makes sense to split an application into multiple deployment units (microservices). Four legitimate technical reasons are outlined: differences in traffic/workload across modules, need for horizontal scaling during peak load, multi-tenancy with data isolation (sharding), and multi-region deployments. Two common but questionable justifications are also challenged: pursuing 100% availability at all costs, and splitting services purely based on storage type. The post argues that microservices should be driven by real technical needs, not cargo cult or buzzword adoption.
