CONCEPTS
Service Level Indicator
A measurable metric that indicates service performance, used to track SLOs.
SLI = (Good Events / Total Events) × 100%
A measurable metric that indicates service performance, used to track SLOs.
The "Yardstick"
SLI (Service Level Indicator) is the specific metric you measure to see if you are meeting your goals. It is "What is happening right now?"
Good vs. Bad SLIs
- Bad (System-Centric): "CPU Usage < 80%." (Users don't care about CPU).
- Good (User-Centric): "Homepage Load Time < 200ms." (Users notice this).
The Formula
Most SLIs are tracked as a ratio of Good Events to Total Events.
SLI = (Number of Successful Requests) / (Total Number of Requests)
If you have 100 requests and 1 error, your SLI is 99%.
ExThe Search Bar
"A team defined their SLI as "Search results returned"."
Impact
The search was returning empty results (0 items found), which counted as "Success". Users were frustrated.
Resolution
They redefined the SLI to "Search results returned containing >0 items within 500ms".
Why SLI Matters
SLIs are the raw metrics (latency, error rate, throughput) that feed into SLO calculations.
Good SLIs are user-focused, not system-focused. Measure what users experience.
SLI vs. Other Metrics
SLI
Measured metric
SLO
Target threshold
SLA
Customer promise
Common Pitfalls
Averages
Avoid average latency. Averages hide outliers. Use Percentiles (p95, p99) instead.
Vanity Metrics
Tracking CPU usage as an SLI. Users do not care about CPU; they care about page load speed.
Related Terms
Frequently Asked Questions
How many SLIs should I have?
Keep it simple. 1-3 SLIs per user journey (e.g., Login, Checkout, Search) is plenty.
What is the most important SLI?
Availability (Did it work?) and Latency (Was it fast?).