Similar to tests that verify business behavior, it can be helpful to test upfront whether an application, or component thereof, is likely to meet their performance expectations in production. The motivation is to prevent major performance drops, potentially caused by introduced errors.

It's important to consider the application logic when constructing performance test scenarios. Some invocations perform more expensive processes than others. Generally it makes sense to construct performance tests after realistic production scenarios, in regard to the frequency and nature of requests.

For example, the ratio of guests browsing an online shop to customers actually performing purchase transactions should somehow reflect the real world.

However, it also makes sense to construct tests that excessively perform expensive invocations, to detect potential issues that emerge when the system is under stress.

In Chapter 9, Monitoring, Performance, and Logging, we will see why performance tests on environments other than production are a poor tool to explore the application's limits and potential bottlenecks. Instead of putting great effort into crafting sophisticated performance test scenarios, it makes more sense to invest into technical insights into production systems.

Still, we will see a few techniques of how to craft simple load tests that put a simulated application under pressure to find evident issues.

A reasonable attempt is to simulate usual load, ramp up the number of concurrent users, and explore at which point the application becomes unresponsive. If the responsiveness breaks sooner than from an earlier test run, this could indicate an issue.