Definition
A Service Level Indicator (SLI) is the concrete metric you measure to judge reliability, such as the percentage of successful checks, request success rate, or response time under a threshold. It's the raw number; the SLO is the target for that number, and the SLA is the promise built on it.
Good SLIs reflect what users actually experience. "Percentage of requests that returned a correct response within 2 seconds" is a far more meaningful SLI than "server CPU usage," because it maps directly to whether the service worked for someone.
Why It Matters
You can't manage what you don't measure. SLIs are the foundation of every reliability conversation — without a well-chosen indicator, your SLO and SLA are meaningless. The right SLI focuses the team on user impact rather than internal vanity metrics.
How It Works
Pick an indicator tied to user experience (uptime, success rate, latency), define exactly how it's measured, and track it continuously. The SLI is then compared to the SLO target over a window. Monitoring tools produce SLIs by recording the outcome of each check or request.
Real-World Example
A team defines its primary SLI as "percentage of checks that returned HTTP 200 within 2 seconds." Over 30 days, 43,150 of 43,200 checks met that bar — an SLI of 99.88%, which they compare against their 99.9% SLO.
Best Practices
- Choose SLIs that map directly to user experience
- Define precisely how each SLI is measured to avoid ambiguity
- Prefer success rate and latency over internal resource metrics
- Measure SLIs continuously, not just during incidents
- Keep the number of headline SLIs small and focused
Common Mistakes
- Picking internal metrics (CPU, memory) that don't reflect user impact
- Leaving the SLI definition vague so results aren't comparable
- Tracking too many SLIs and losing focus
- Measuring SLIs with sampling too coarse to be accurate
- Confusing the SLI (the measurement) with the SLO (the target)
In Monitoristic
Your uptime percentage and response-time data in Monitoristic are exactly the kind of SLIs you'd track. Each check is recorded as a success or failure against the expected status code, producing a clean, user-focused indicator.