Monitoring, Alerting & SLI/SLO/SLA

Observability gives you signals; now someone has to watch them and react. Monitoring watches the signals automatically and alerts a human when something's wrong. And to know what "wrong" even means, teams set clear reliability targets using three linked terms: SLI, SLO, and SLA.

These three acronyms sound bureaucratic but are genuinely useful: they turn the vague question "is it working?" into something specific and measurable that a whole team can agree on.

A home thermostat is the picture: it monitors the temperature, has a target you set (the goal), and a landlord's lease might even guarantee a range (the promise). Three layers — measure, target, promise — exactly like what follows.

Monitoring and Alerting

Monitoring is automatically watching the metrics and logs from the last topic, continuously. Alerting is the part that pages a human when a signal crosses a line you care about — error rate too high, response too slow, the system down. The goal is to learn about problems from your monitoring before users do, so you're responding to an alert at 2 AM, not to angry customers in the morning.

SLI: What You Measure

An SLI (Service Level Indicator) is the actual measurement of how the service is doing — a real number, like "the percentage of requests that respond in under one second" or "the percentage of reminders that fire on time". It's the concrete, measured reality. Everything else builds on it: you can't have a target or a promise without first deciding what you're measuring.

SLO: The Target

An SLO (Service Level Objective) is the internal goal for that indicator — the target you're aiming for, like "99.9% of requests under one second". It's the line that separates "healthy" from "needs attention", and it's what alerts are set against. Crucially, SLOs are deliberately not 100%, because perfect reliability is impossibly expensive; a good SLO picks a sensible bar that's worth the cost.

SLA: The Promise

An SLA (Service Level Agreement) is the formal promise made to customers — a contractual commitment, often with consequences (like refunds) if it's broken. An SLA is usually set looser than the internal SLO, so the team has a safety margin: they aim for the tougher internal target and only breach the customer promise if things go badly wrong. Indicator, objective, agreement — measure, goal, promise.

Cadence's reliability has all three. Their SLI is "the percentage of reminders that fire within one minute of the chosen time", measured continuously. Their SLO is "99% of reminders within one minute" — the internal target their monitoring checks, paging whoever's on call if it dips. They don't offer a contractual SLA yet, but if they sold a premium plan, that promise would sit a little looser than the 99% they aim for internally, giving them a margin.