DevOps, SRE & Cloud Interview · Observability & SRE · Lesson 2 of 3
SLIs, SLOs & error budgets
SRE turns 'reliability' into numbers you can manage. The chain is SLI → SLO → error budget.
- SLI (Indicator): a measured signal of user-facing health — e.g. the fraction of requests served under 300ms, or the success rate.
- SLO (Objective): the target for an SLI over a window — e.g. '99.9 percent of requests succeed over 30 days'. It defines 'good enough'.
- Error budget: 100% minus the SLO — the allowed unreliability. 99.9 percent leaves ~43 minutes/month of budget to spend on risk, deploys, and experiments.
- SLA: a contractual promise to customers with penalties — usually looser than the internal SLO you run to.
The error budget is a decision toolThe budget aligns dev and ops: budget remaining → ship features freely; budget exhausted → freeze risky changes and pour effort into reliability. It replaces arguing about '100 percent uptime' with a shared, quantitative rule.
100% is the wrong targetChasing 100 percent reliability is infinitely expensive and blocks all change. The point of an SLO is to pick a target that keeps users happy while deliberately spending the remaining budget on velocity.
◆ Lock it in
- SLI = measured health signal; SLO = target for it; error budget = 100% minus the SLO.
- The error budget is a decision rule: budget left → ship; budget gone → freeze and fix reliability.
- SLA is the external contract (with penalties); the internal SLO is usually tighter. 100% is never the goal.
Feynman drill — say it out loudExplain how an error budget is computed from an SLO and how a team uses it to decide whether to ship.
Step 1 rate your confidence · Step 2 pick your answer
A service has a 99.9% availability SLO. What is its error budget?
Step 1 — how sure are you?