An SLA (Service Level Agreement) is a documented or contractual commitment that specifies the minimum performance standards a service provider will maintain. It typically covers metrics like monthly uptime percentage, response latency, error rates, and support ticket resolution times.
How it works: The provider sets target metrics (for example, 99.9% uptime means no more than 43 minutes of downtime per month). Monitoring systems continuously sample the service's availability and performance. At the end of each billing period, the total downtime is calculated as a percentage of total time. If the provider falls below the committed threshold, the customer receives a remedy—usually a service credit applied to the next bill, not a cash refund.
Example: A cloud hosting provider commits to 99.95% monthly uptime. If actual uptime falls to 99.0%, the customer receives a 10% credit on that month's bill.
Metrics commonly in SLAs:
| Metric | Example |
|---|---|
| Uptime | 99.9% monthly availability |
| Response time | API responses within 200ms (p95) |
| Support response | Critical tickets answered within 1 hour |
| Error rate | Fewer than 0.1% failed requests |
WarningUptime is measured from the provider's monitoring infrastructure, not from the customer's end-to-end experience. DNS failures, CDN issues, or client-side network problems typically do not count as SLA breaches. Also, high uptime percentages still allow significant cumulative downtime: 99.9% = ~8.76 hours per year.
Use /features/monitoring to track your own infrastructure against SLA targets, or /tools/uptime-calculator to calculate downtime budgets from a percentage commitment.