Uptime SLA99.99 %, and we pay it back ourselves.
Every SLA promises a number. The part that matters is who has to notice it was missed, and almost everywhere that is you — through a form, inside a window, with evidence. Here the credit is computed from the same incident record that drives the status page, and applied to your balance without being asked.
The commitment
What 99.99 % actually buys you.
A percentage is easy to print and hard to feel. Here is the same commitment expressed as the thing you would actually experience: minutes in which your server could be unreachable before we owe you money.
Allowance a day
9 seconds
Allowance a week
1 minute
Allowance a month
4.3 minutes
Allowance a year
52.6 minutes
Measured per region, on network reachability — whether packets reached your server, not whether it felt quick. Degradation is recorded on the status page and does not silently count as uptime, but it is not what this figure is about. The current estate average is 99.9971 %.
What a bad month is worth
As a percentage of the monthly rate for the affected service, credited to your balance. It compounds with nothing and expires never.
-
Below 99.99 %over 4.3 min down in a 30-day month10 % of the monthly rate5,00$ to your balance
-
Below 99.90 %over 43.2 min down in a 30-day month25 % of the monthly rate12,50$ to your balance
-
Below 99.50 %over 216 min down in a 30-day month50 % of the monthly rate25,00$ to your balance
-
Below 99.00 %over 432 min down in a 30-day month100 % of the monthly rate50,00$ to your balance
Three things an SLA usually hides
The number is the easy part. The process is where they get you.
Most agreements are honest about the percentage and quiet about everything after it: a claim you have to file, a window you have to file it in, evidence you have to supply, and a credit that only exists against a future invoice you may never send. None of that is here, and the reason is simple — we already have the record, so making you prove it would only be a way of hoping you do not bother.
The record this is computed from- No claim to file The credit is computed from our own incident log and applied. There is no form.
- No window to miss Nothing expires because you were busy in the thirty days after an outage.
- No credit you can only spend here It lands on your balance. Cancel afterwards and the balance is still yours.
What it does not cover
Short, and stated up front rather than in a schedule at the end. An exclusion you find after an outage is a term you were not offered.
- Maintenance announced at least 72 hours ahead, inside a window you were told about.
- Anything caused by your own configuration, software or resource exhaustion.
- Suspension under the acceptable use policy, which is a decision rather than a failure.
- An attack we absorbed that degraded your service without our network losing reachability.
- Failure of a network you reach us through that is not ours.
Everything else counts, including our own mistakes, a failure at a facility we chose, and an attack that took a region off the internet. Choosing a supplier who failed is our decision to have made, not a force of nature to shelter behind.
About the agreement.
How is the credit actually applied?
As a line on your ledger at the end of the month, computed from the incident record for the regions your services sit in. It behaves like any other balance: it pays for renewals, it does not expire, and it is not clawed back if you cancel.
Do I have to claim it?
No, and there is no form to find. We already hold the record the credit is computed from, so asking you to prove an outage we recorded ourselves would only be a way of hoping you never get round to it. Signed in, this page shows you what is owed today.
What exactly is being measured?
Network reachability of the region your service sits in, over a rolling window, from the incident log. Only minutes in which traffic could not reach it count. An event that made things slow without breaking reachability is published on the status page and does not reduce the percentage.
Why is the window rolling rather than calendar?
Because a calendar month lets an outage on the 31st cost nothing on the 1st. A rolling window means an incident carries its weight for a full month wherever it lands, which is worse for us and more honest to you.
What if a region is down for days?
Below 99 % the credit is the entire monthly rate for the affected service — you pay nothing for that month. Beyond that we would be talking about migration and a refund rather than a credit, and an engineer would be talking to you directly long before you read this page.
Does the SLA apply to bare metal and colocation too?
The network commitment applies to everything we route, including colocated hardware. Hardware failure on a dedicated server is covered separately: failed drives, memory and power supplies are replaced within two hours of diagnosis, around the clock, at no charge.