Uptime Calculator — SLA Downtime Calculator
#What this tool does
Enter an uptime percentage and get the exact downtime budget broken into year, month, week, and day. Useful for evaluating SLA tiers, planning maintenance windows, and setting alerting thresholds. Calculations assume a 365-day year and 30-day month.
#Common SLA tiers
These are the standard uptime tiers used across the industry. The "nines" shorthand refers to the number of 9s after the decimal point.
| Uptime % | Allowed downtime/year | Shorthand |
|---|---|---|
| 99% | 3d 15h 39m (87.6h) | "two nines" |
| 99.5% | 1d 19h 49m (43.8h) | — |
| 99.9% | 8h 45m 36s | "three nines" |
| 99.95% | 4h 22m 48s | — |
| 99.99% | 52m 35s | "four nines" |
| 99.999% | 5m 15s | "five nines" |
Each additional nine cuts allowed downtime by roughly 10x. Moving from 99.9% to 99.99% means your entire annual downtime budget drops from nearly 9 hours to under 53 minutes.
#What counts as downtime?
SLA definitions vary, but most providers measure downtime as any period where the service is unavailable or returns errors above a defined threshold.
Full outages — the service is unreachable. HTTP requests time out or return 5xx errors. This always counts.
Degraded performance — response times exceed SLA thresholds (e.g., p99 latency above 10s). Some SLAs count this as partial downtime; others ignore it entirely.
Partial outages — a subset of regions, endpoints, or users are affected. Cloud providers typically measure this as an "error rate" rather than binary up/down.
Scheduled maintenance — most SLAs explicitly exclude planned maintenance windows from downtime calculations. Check the fine print. GCP, AWS, and Azure all exclude scheduled maintenance from their SLA metrics.
Unexpected application crashes — like OOM kills or processes hitting exit code 137 — count against your uptime if they cause user-facing errors.
#SLAs from major cloud providers
Cloud provider SLAs define financial credits when uptime falls below the guaranteed threshold. They do not guarantee the service will never go down — they guarantee compensation when it does.
Google Cloud — Compute Engine offers 99.99% for multi-zone deployments (single instance: 99.5%). GKE Autopilot provides a 99.95% SLA for the regional cluster control plane. Cloud SQL offers 99.95% for regional instances. (Source: Google Cloud SLA page, March 2026)
AWS — EC2 offers 99.99% for multi-AZ deployments (single instance: 99.5%). RDS Multi-AZ provides 99.95%. EKS control plane SLA is 99.95%. (Source: AWS Service Level Agreements, March 2026)
Azure — Virtual Machines offer 99.95% for availability sets, 99.99% for availability zones. Single instances with premium SSD get 99.9%. AKS control plane provides 99.95% with the Uptime SLA feature enabled. (Source: Azure SLA summary, March 2026)
When you deploy containerized applications, your effective uptime depends on both the underlying infrastructure SLA and your own application reliability. A 99.99% compute SLA means nothing if your application crashes every hour. For help choosing a platform that meets your SLA needs, see our cloud hosting platform comparison.
#Open source
This tool is powered by @azin-tech/mini-tools, an open-source developer toolkit. View source on GitHub.
Auto-deploy into your own cloud
Push code, AZIN handles the rest. Auto-detected builds, your cloud account, no vendor lock-in.