Load balancing and scaling
Load balancing spreads requests across several servers and keeps applications available during traffic spikes and server failures.
Why a single server is a risk
Load balancing stops a single server from becoming the point of failure for the whole application. If that server fails or is overloaded, the application stops. For shops, portals and platforms with many users, that means lost revenue or blocked business processes.
A load balancer distributes incoming requests across several servers. If one fails, the others take over.
Scaling: vertical or horizontal
Vertical: a larger server provides more computing power. It is simple, but it has limits.
Horizontal: several servers share the load. This is more flexible and more resilient, but takes more effort to operate.
In the cloud, the number of servers can adjust to the load automatically. This saves costs in quiet periods. It does, however, require an application designed for it.
Prerequisites in the application
For several servers to work together, the application has to be prepared for it. Sessions and cache are held centrally, for example in Redis. Uploaded files are kept in shared storage. The database is replicated where needed. Regular health checks automatically remove faulty servers from the pool.
Approach
First, we check whether caching and optimisation already solve the problem, which is often cheaper than extra servers. If not, we plan the architecture for horizontal scaling and introduce it step by step.
Before campaigns or expected traffic spikes, we run load tests. They show in advance how many simultaneous users the system can handle and where the next bottleneck lies.