Your e-commerce platform handles 1,000 requests per second on a normal day. Then a major sale starts, and demand jumps far beyond the capacity you normally run. App servers queue requests, the database connection pool fills up, and customers start seeing slow pages or errors.

This is the challenge of traffic spikes: demand rises faster than the system can safely absorb it.

A common mistake is treating every spike the same way. A read-heavy news spike, a flash-sale checkout spike, a retry storm, and a DDoS event need different responses. The right answer depends on what is spiking, whether the work is read or write heavy, whether users need an immediate answer, and what can be degraded safely.

A strong system design answer goes past "auto-scale and add cache." It explains how to survive the scaling lag, protect the bottleneck, shed or queue excess work, and preserve the most important user flows.