Load balancing and scaling
Load balancing spreads requests across several servers and keeps applications available during traffic spikes and server failures.
The single-server risk
If a single server fails or is overloaded, the whole application stops. For shops and portals, that means lost revenue or blocked business processes.
Scaling: vertical or horizontal
Vertical: a larger server provides more computing power, but it has limits.
Horizontal: several servers share the load. This is more resilient but takes more effort to operate.
In the cloud, the number of servers can follow the load automatically if the application is designed for it.
Application prerequisites
Sessions and cache are held centrally, for example in Redis, and uploads in shared storage. The database is replicated where needed. Health checks remove faulty servers from the pool automatically.
Approach
Caching and optimisation are often enough, and cheaper than extra servers. If not, we plan the architecture for horizontal scaling and introduce it step by step.
Load tests before campaigns show how many simultaneous users the system can handle.