SLI (Service Level Indicator)
An SLI is a measurement of some aspect of service quality, such as the percentage of successful requests.
A Service Level Indicator is a quantitative measure of service quality that reflects user experience — most often availability (the share of successful requests), latency (the share served under a threshold), or error rate. It is the raw signal: a number, measured over time, that an SLO sets a target against and an SLA promises a level of.
Why it matters: SLOs and SLAs are only as trustworthy as the indicators beneath them. If the SLI is poorly chosen or measured from the wrong vantage point, every target built on it is fiction. A good SLI is expressed as a ratio of good events to total events, which makes it easy to aggregate, compare across windows, and reason about as a percentage.
A common misconception is that any metric is an SLI. Server-side CPU or memory usage are useful health metrics but they are not SLIs, because they don't directly describe what a user experiences. A true SLI answers 'is the service doing its job for the people using it?' — which usually means measuring from the user's perspective, ideally from outside your own infrastructure.
Choosing the right vantage point matters as much as the metric. An availability SLI measured only from inside your data center can look perfect while users on the public internet see failures. External, multi-region checks give an SLI that reflects reality rather than your own optimism.
Frequently asked questions
- What are common examples of SLIs?
- The three most common are availability (percentage of successful requests), latency (percentage of requests served under a target time), and error rate. Each is expressed as a ratio of good events to total events.
- What makes a good SLI?
- A good SLI closely tracks user experience, is measured as a good-events-over-total ratio, and is observed from a vantage point that reflects how users actually reach the service — often externally, from multiple regions.
- Is a health metric like CPU usage an SLI?
- No. Resource metrics help diagnose problems but don't describe user experience. An SLI must measure something a user would notice, such as whether their request succeeded or how long it took.
Start monitoring in under a minute
Free forever, no credit card. Upgrade when you need finer intervals.