“Uptime” and “availability” are used interchangeably in most conversations. But they measure different things, and confusing them can lead to misleading SLA reports and missed incidents.
Uptime: Is the System Running?
Uptime measures whether a system is powered on and its processes are active. A server with 365 days of uptime has been running continuously without a reboot for a full year.
Check server uptime on Linux:
$ uptime14:23:45 up 365 days, 0:00Uptime says nothing about whether users can actually use the service. The server can be “up” while:
- The application has crashed
- The database is unresponsive
- Network connectivity is lost
- The service returns errors for every request
Availability: Can Users Use It?
Availability measures whether the service is accessible and functioning correctly from a user’s perspective. A service is available when it responds correctly to user requests within acceptable time limits.
Availability is what SLAs should measure, because it reflects actual user experience. The formula:
Availability = Successful requests / Total requests x 100Or time-based:
Availability = (Total time - Downtime) / Total time x 100Use the uptime calculator to convert between availability percentage and downtime duration.
Why the Distinction Matters
Scenario 1: Server Up, Service Down
Your Kubernetes node has been running for 90 days (100% uptime). But a misconfigured deployment caused your API to crash-loop for 2 hours. Server uptime: 100%. Service availability: 99.9%.
Scenario 2: Server Rebooted, Zero Downtime
You do a rolling restart of your 3-node cluster. Each node reboots (technically 0% uptime for ~30 seconds each). But because the load balancer routes traffic to healthy nodes, users experience zero interruption. Server uptime: ~99.99%. Service availability: 100%.
Scenario 3: Partial Degradation
Your service is running but returning errors for 20% of requests for 30 minutes. Server uptime: 100%. Service availability: debatable. Is it “up” if 1 in 5 requests fails?
What to Measure in Your SLA
SLAs should define availability, not uptime. A well-written SLA specifies:
- What “available” means — “Returns HTTP 200 within 2 seconds for health check endpoint”
- How it’s measured — “Measured by external monitoring checks every 30 seconds from 3+ regions”
- What’s excluded — “Scheduled maintenance windows (announced 48h in advance) are excluded”
- The target — “99.9% monthly availability”
- Consequences — “Below 99.9%: 10% service credit. Below 99.0%: 30% service credit”
How to Measure Availability
External Synthetic Monitoring (Best)
An uptime monitoring tool sends requests from outside your network. This measures the same thing users experience: DNS resolution, network path, application response.
Request Success Rate
Track the ratio of successful responses (2xx) to total responses from your access logs or load balancer metrics. This uses real traffic data.
Real User Monitoring (RUM)
JavaScript agents in your application report performance and errors from actual user sessions. This captures the real user experience but doesn’t work during zero-traffic periods.
The Right Combination
Use external monitoring as the SLA measurement source (it catches everything), supplement with request success rates for real-time dashboards, and RUM for user experience insights.
Partial Availability
Real systems don’t just go from “100% up” to “100% down.” Partial failures are common:
- One API endpoint fails while others work
- Mobile app works but web dashboard doesn’t
- US region healthy, EU region degraded
- Read operations succeed, write operations fail
Your monitoring should detect these partial failures. This is why monitoring multiple endpoints matters more than just checking your homepage.
Summary
| Aspect | Uptime | Availability |
|---|---|---|
| Measures | System is running | Users can use the service |
| Perspective | Internal (server) | External (user) |
| Detected by | System commands | External monitoring |
| SLA metric? | No | Yes |
| Captures app crashes? | No | Yes |
| Captures network issues? | No | Yes |
For SLAs and monitoring, focus on availability. It’s what your users care about and what your business depends on.
Related tools:
- Uptime Calculator — Convert availability percentages to downtime
- Error Budget Calculator — Track your reliability budget