What is the difference between load testing and stress testing?

Load testing answers "does the system meet our SLOs at the traffic we actually expect?" You ramp to your target concurrent users (or RPS), hold there for a sustained period, and measure throughput, latency percentiles, error rate, and resource utilisation. Pass/fail is judged against the service-level objectives (e.g. p95 < 500ms, error rate < 0.1%).

Stress testing answers "what happens when we exceed expected load — does the system degrade gracefully or fall over?" You push past target capacity until something breaks: requests start timing out, the database connection pool saturates, a downstream service rate-limits you. The valuable output isn't a single number but a description of how the system fails — does it recover after the spike subsides? Do circuit breakers trip? Do queues drain in reasonable time?

Two related types worth knowing: soak testing (sustained moderate load over hours/days to find memory leaks, log-disk fill-up, slow leaks of any kind) and spike testing (sudden 10x bursts simulating a marketing campaign or DDoS).

In an interview, the key signal is that you understand all four answer different production risks and that you'd never substitute one for another.