Designing a Scalable Architecture for Million‑Level DAU Systems

Recent incidents of service outages caused by insufficient scalability highlight the need for a robust architecture that can automatically expand and contract under sudden traffic spikes.

The request flow starts with DNS resolution, which directs users to the appropriate regional data center, followed by L4 load balancing (typically LVS) for traffic forwarding and L7 load balancing (commonly Nginx) for application‑level routing, authentication, logging, and monitoring.

Backend services can be deployed as a monolithic application for simple or small‑team projects, or as microservices when the number of interfaces and team size grows, allowing independent development, packaging, and deployment.

Caching layers (e.g., Redis, Memcached) reduce database latency, while the database layer must support high availability through master‑slave replication, sharding by time or user ID, and partitioning to handle terabyte‑scale data.

For traffic beyond the capacity of private data centers, a hybrid‑cloud architecture distributes load between private and public clouds, requiring network interconnectivity (often via dedicated lines) and careful placement of services across clouds.

Full‑link elastic scaling ensures that L4/L7, services, cache, and databases can dynamically scale; services should be stateless and use metrics‑driven auto‑scaling, while cache and database expansions must account for warm‑up times and data synchronization delays.

A three‑level degradation strategy (invisible, user‑perceptible, and severe) provides a safety net by gracefully shedding load when resources become constrained.

Additional supporting mechanisms such as decision‑support systems and alerting are essential for maintaining reliability at the ten‑million‑plus DAU scale.