Availability
What we promise about uptime and what a missed month gives back.
The commitment
Measured over a calendar month, for every company on a paid plan.
| Service | Monthly availability |
|---|---|
| Mail delivery, inbound and outbound | 99.9 percent |
| Inbox, Chat, Cloud, Calendar, Live, Book and Console | 99.9 percent |
| Meeting media in Live | 99.5 percent |
Announced maintenance is excluded: at most four hours a month, announced to your administrators two days ahead and never between 08:00 and 20:00 Singapore time.
How it is measured
A probe outside our own infrastructure asks each service every five minutes. Availability is the share of those windows in which it answered. A window counts as down only after two failed probes in a row, so one slow answer is not an outage and we cannot quietly round a bad month up.
Mail counts as available when a message is accepted for delivery and a queued message is relayed. Each product host counts as available when its health check answers within two seconds.
Service credits
When a month falls short, a company on a paid plan may ask for a credit against its next invoice within thirty days of the month ending.
| Availability in the month | Credit |
|---|---|
| Below 99.9 and at least 99.0 percent | 10 percent of that month |
| Below 99.0 and at least 95.0 percent | 25 percent of that month |
| Below 95.0 percent | 50 percent of that month |
Credits are the remedy for a missed month. They do not apply to an outage caused by the company's own network, its own identity provider, a supplier of its own, or use outside the agreement.
What is not covered
Trials. Features marked beta in the product. Outages of the wider internet, of a company's own connection or of a local exchange node it opted into, which is a cache rather than the main path. Events outside anyone's control.
Status and incidents
status.flo.now shows the current state of every host from the same probe, with the last twenty four hours of checks. It is hosted away from our own infrastructure, so it stays up when we are down. The full probe record behind a month's figure is sent to any company that asks for it.
When an incident affects your company, your administrators hear from us within fifteen minutes of detection and at least hourly until it is resolved. Anything longer than thirty minutes gets a written account of what happened and its cause, within five working days.
Recovery targets behind the commitment
The database can be restored to any point in the last fifteen minutes. A nightly copy is kept thirty five days under a lock that stops anyone deleting it early. A full restore from that copy targets four hours and is rehearsed every quarter rather than assumed.
Questions about any of this go to support@fastflowup.com.