Uptime

SLA

Also called service level agreement.

SLA is a written promise to customers about service levels, usually with credits or refunds if you miss it. It is a contract, not a goal.

How it is measured

Read the clauses: which metric (uptime percentage), how it is measured and by whom, the window (monthly), exclusions (maintenance, force majeure), and the remedy (10 percent credit below 99.9, 25 percent below 99). Calculate each month from your own records.

Customers usually must claim within 30 days with evidence. Keep probe logs so you can answer a claim in an hour.

Worked example

A hosted forms tool promises 99.9 percent monthly, with a 10 percent credit below that and 25 percent below 99.0. In June the vendor has one 63-minute outage. June has 43,200 minutes, so 63 bad ones give 99.854 percent.

That falls below 99.9 and above 99, so the credit is 10 percent of a 49 dollar monthly plan, 4.90 dollars. The customer files within the window and finance applies it. The internal SLO was 99.95, so the team had already treated the month as a miss.

How it differs

SLA is the external promise with consequences. SLO is the internal target, usually stricter. The SLA excludes the margin you manage to. The SLO excludes the legal remedy.

Common errors

Promising more than your dependencies provide. Vague measurement. No exclusions defined. Making the SLA equal to the SLO. Forgetting the credit process. Quoting it on a page without the fine print.

In practice

Set your internal SLO above your SLA so you have margin before money is owed. Read the contract clause on measurement. Archive probe logs for at least a year.

See also

SLO, SLI

Sources

Count this on a real site.

Watch my website