---
name: Reliability / SLOs
description: Define and defend service-level objectives.
---
# Reliability & SLOs

An SLO is a measurable reliability target ("99.9% of requests succeed within 200ms over 28 days"). It converts "is the service up?" from opinion to a number — and bounds how much risk you take.

## Method
- **Set SLIs first.** Service-level indicators: the metrics (error rate, latency).
- **Set SLOs.** The target values, over a window.
- **Error budget.** 99.9% over a month = ~43 minutes of allowed unavailability. As long as you're under budget, you can ship features; when you burn it fast, freeze and stabilize.

## Why it matters
- SLOs make the reliability/velocity trade-off explicit and shared with the business. "We can't launch because we burned the error budget" is a fact, not an argument.
- 100% is the wrong target. The cost of the last 0.1% exceeds the value for almost any product.