Uptime & SLA

Uptime is the share of time a service stays reachable, and the percentages are worth converting into minutes before you let them impress you. A 30-day month holds 43,200 minutes. A promise of 99.9% concedes a tenth of one per cent of that, so 43.2 minutes may go missing every month without anything being breached. Spread across a year of 8,760 hours, the same 99.9% permits 8.76 hours of darkness. Adding a nine shrinks the allowance by a factor of ten: 99.99% leaves 4.32 minutes in a month and roughly 53 minutes across a year. Removing one blows it open: a flat 99% tolerates 7.2 hours a month and 87.6 hours a year, which is more than three and a half days. Each additional nine also costs disproportionately more to engineer, which is the real reason the cheap tiers stop where they do.

That arithmetic is the sole portion of the document nobody can argue with. Everything else lives in the definitions. Scheduled maintenance is normally carved out entirely, the clock frequently starts when you open a ticket rather than when the service stopped responding, the provider's own monitoring is the evidence of record, and the quantity being measured is the platform rather than your site. Your site additionally depends on your application, your DNS, your database and whatever you deployed last Friday, none of which the agreement has any view on. So read the exclusions first and the headline second. Then read the remedy, which is where the document quietly changes subject: a breach yields credit against a future invoice, proportional to fees for the affected period, and you generally have to claim it yourself inside a stated window. Since an hour offline costs most businesses considerably more than an hour of hosting, the compensation is a gesture. Availability compounds downwards, too: a shop leaning on a payment gateway, a mail relay and a headless CMS inherits the frailty of all three, and none of those vendors have any agreement with your customers.