Opening Scenario
Your team just shipped a new feature. Load testing looks great — your service handles 10,000 requests per second. But users are complaining that the app "feels slow." Your p99 latency is 4 seconds. A colleague says "we should add more servers." Another says "we need to optimize the hot path." Both might be right — but for different reasons, and fixing the wrong one wastes weeks.
Understanding what you're actually measuring is the prerequisite to fixing it.
Question Progression
1. What are you actually optimizing?
Question: You have two metrics on your dashboard: "requests per second" (RPS) and "p99 latency." Your users are unhappy. Which metric tells you what's wrong?
Reasoning: RPS tells you how much work your system completes over time. p99 latency tells you how long the slowest 1% of requests take to complete. These measure different things. A system can have high RPS and terrible…
In the full tutorial
- When to Reach for This