Choose a definition before calculating a percentage
Uptime can mean powered on, locally hashing, connected to a pool, or producing accepted work. Those states can differ. A useful availability report names the state, time window, measurement source, and how planned maintenance or missing data is treated.
Track causes as well as durations. Electrical interruptions, cooling faults, maintenance, network failures, and pool issues need different responses. Counting every unavailable interval as hardware failure obscures what improvements would actually restore productive work.
Step-by-step
- Define the metric. State the productive condition and denominator period. Decide how planned maintenance and unknown intervals are classified, without silently hiding them.
- Collect event intervals. Use consistent timestamps, note overlapping events, and avoid double-counting time when several alarms describe the same outage.
- Report causes and energy. Show lost productive time by category and reconcile with power states. Use the report to prioritize practical corrective action.
Worked example
Illustrative example. A hypothetical worker produces accepted work for 22 hours in a 24-hour observation window. Its defined productive availability is 22 ÷ 24 × 100, approximately 91.7%, even if it stayed powered all day.
The percentage describes that window and definition only. It does not establish annual reliability or imply that energy use fell by the same proportion.
Common mistakes
- Calling powered-on time productive uptime.
- Excluding maintenance without stating the policy.
- Adding overlapping alarm durations into an inflated downtime total.
Questions and answers
Should unknown data count as uptime?
Do not silently assume it. Report the gap and its treatment separately. A data outage can prevent you from confirming productive operation even if the machine may have continued running.
Apply this to your setup
Publish the metric definition beside productive hours, denominator hours, planned maintenance, overlapping incidents, and unknown-data intervals. Keep powered availability and accepted-work availability distinct. When totals do not reconcile to the observation window, resolve overlaps or missing classifications before using the percentage in a revenue scenario.
Sources and next steps
Read uptime and energy and monitoring baselines.
How this guide was prepared: AI-generated drafting with the linked technical references and illustrative calculations. It does not report first-hand equipment tests or a named specialist review.