SLA vs SLO vs SLI: what is the difference?
These terms get mixed up constantly. The easiest way to remember them is as a chain: you measure an SLI, you aim for an SLO, and you promise an SLA.
SLI — the measurement
A Service Level Indicator is a number you can measure. Typical examples are the share of successful requests, the share of requests faster than 300 ms, or the share of minutes in which the site answered a health check.
SLO — the internal goal
A Service Level Objective is the target you set for an SLI. “99.9% of requests succeed over 30 days” is an SLO. It is a goal for your team, tighter than anything you promise customers.
SLA — the contract
A Service Level Agreement is the external promise, usually with consequences such as service credits if you miss it. The SLA should be looser than the SLO so that you have room to react before a breach.
A worked example
- SLI: successful responses ÷ total responses
- SLO: 99.95% over a rolling 30 days
- SLA: 99.9% per month, with 10% credit below that
With the 99.95% SLO you have an error budget of about 22 minutes per month. Once it is spent, you slow down risky changes and focus on stability.
Use the calculator on the home page to turn any of these percentages into minutes.
Measure the real number
A calculator tells you the target. Upzilla tells you what you actually delivered.