Uptime (Availability)
Uptime (Availability): The percentage of time the service is operational.
Formula
Uptime = (Total time - Downtime) / Total time × 100Computable definition
The canonical formula in the GASP formula grammar. The MCP server compiles it to SQL over the datasets listed below.
1 - sum(field.downtime_minutes) / (period_days * 1440)Unit: percent (the formula yields an unscaled ratio)
Notes: Total time is period_days * 1440 minutes. Scheduled maintenance counts as downtime (engineering.md).
Data requirements
Grain: Per service per interval, aggregate per period · Frequency: Real-time · Aggregation: aggregate (month) · Staged: no · Main source: Monitoring.
| Field | Dataset | Type |
|---|---|---|
downtime_minutes | service_metrics | integer |
period_start | service_metrics | date |
Example SQL (Postgres)
SELECT (1 - (CAST((SELECT COALESCE(SUM(downtime_minutes), 0) FROM service_metrics WHERE period_start >= :period_start AND period_start <= :period_end) AS DOUBLE PRECISION) / NULLIF(((:period_end - :period_start + 1) * 1440), 0))) AS uptime Access: recommended
sensitivity tier internal,
access scope unscoped.
Guidance your policy engine enforces.
From an agent: with the GASP MCP server, call
lookup_metric({ "name": "Uptime (Availability)" }) or generate_query({ "metric": "Uptime (Availability)", "adapter": "<your adapter JSON>" }).
Benchmarks
- 99.0%: ~7.3 hours downtime (unacceptable for most SaaS)
- 99.9%: ~43 minutes downtime (three nines)
- 99.95%: ~22 minutes downtime (typical SLA)
- 99.99%: ~4.3 minutes downtime (four nines, high reliability)
What It Tells You
Basic platform reliability.
Common Mistakes
- Excluding "scheduled maintenance" (customers don't care why it's down)
- Measuring only core service, not dependencies
- Not weighting by traffic/usage
Related Metrics
Connected in the GASP relationship graph.
Upstream: what drives this
- MTTR reduces Uptime (Engineering)