Every hosting provider advertises uptime. Almost none of them define it, and fewer still attach a consequence to missing it. OneMiners publishes 30-day rolling uptime for each individual machine, excludes up to 24 hours of scheduled maintenance a month, and compensates you under contract when a miner falls below 95%. Here is why each of those three clauses matters.
"99% uptime" is one of the least informative claims in hosting. It does not say whether the figure is measured per facility or per machine. It does not say what counts as downtime. It does not say over what window. And it almost never says what happens if the number is missed.
Those four omissions are where the entire value of an uptime promise lives.
Key takeaways
- Uptime is reported per miner, on a 30-day rolling window, on the Workers page — not as one facility-wide average.
- Up to 24 hours of scheduled maintenance per month is excluded from the calculation, and the same exclusion applies to the dashboard's overall figure.
- Below 95%, compensation is due under the contract. An information button in the app states the allowance.
- Average uptime across the network runs at 98%+, which is the operating reality; 95% is the contractual floor, which is the promise.
- A facility-level average can hide a failed machine indefinitely. A per-machine figure cannot.
First, the stakes: why an average hides the only thing you care about
Imagine a site with a thousand miners, ten of which have failed and sat offline for a month. Facility uptime reads 99%. The operator's marketing claim is true. And if one of those ten machines is yours, your uptime for the month was zero.
This is not a hypothetical edge case, it is the normal failure mode of aggregate reporting. Hashboards fail, power supplies fail, a fan seizes and a unit throttles itself into uselessness. In a fleet of any size, some small number of machines is always in trouble. An average is mathematically guaranteed to hide them.
Per-machine reporting inverts the incentive. If every client can see their own unit's rolling figure, a failed miner becomes visible to exactly the person who cares most, on the day it happens rather than at the end of a quarter.
The three clauses that make an uptime number mean something
1. The unit of measurement
Our Workers page reports a 30-day rolling percentage for each machine you own, next to that machine's live status. The dashboard also carries an overall figure, but the per-machine number is the one that is contractually meaningful.
2. What counts as downtime
Up to 24 hours per month of scheduled maintenance is excluded from the calculation. That cap is the important half of the sentence. An uncapped maintenance exclusion is an unlimited licence to reclassify downtime, and a provider who will not tell you their cap has told you something.
3. The remedy
This is the clause almost nobody publishes. A 99% target with no consequence for missing it costs the provider nothing to promise. A 95% floor with compensation attached costs the provider real capital every time it is breached, which is precisely why it changes behaviour. The allowance is stated in-app behind an information button, so it is visible to the client rather than buried in a signed PDF.
Why the floor is lower than the average, and why that is the honest way round
Our network averages 98%+ uptime. The contractual floor is 95%. Those two numbers being different is not a contradiction, it is what a real commitment looks like.
An operator who sets a contractual floor equal to their best-case average is either going to breach it routinely or define downtime loosely enough that they never do. Setting the floor below the operating average leaves genuine headroom for the physical world — a grid event, a failed transformer, a hardware batch with a defect — while still being a number with teeth. The average tells you how it actually runs. The floor tells you what you are owed when it does not.
What to ask any hosting provider
- Is your uptime figure per machine or per facility? If per facility, your individual miner's performance is invisible.
- Over what window? A rolling 30 days behaves very differently from an annual average, which can absorb a two-week outage without moving.
- What is excluded, and is the exclusion capped? Uncapped scheduled maintenance makes the headline meaningless.
- What is the remedy below the threshold, and is it in the contract? If the answer is goodwill, there is no threshold.
- Can I see the figure myself, without asking? A number you have to request is a number someone can edit.
Red flags
- An uptime claim with no stated measurement window.
- "Best effort" or "target" language anywhere near the percentage.
- A facility-level SLA presented as though it covered your machine.
- No in-app visibility, only a monthly report from the provider.
- A maintenance exclusion with no cap.
Final thoughts
Uptime is the one metric in hosting where the marketing number and the useful number have drifted furthest apart. Almost every provider can truthfully claim something in the high nineties, because almost every provider is measuring the thing that makes that claim easy.
The useful question is not how high the number is. It is who it is measured for, what has been taken out of it, and what happens when it is missed.
Hardware is a purchase. Uptime is the product.
Three steps, in order
Check the per-miner figure → read the maintenance cap → confirm the remedy in writing.
See the hosting locations
