Uptime & SLA
Downtime allowed per day, week, month and year for an availability figure, the availability a measured outage gives, and the uptime of chained components.
The permitted outage for any number of nines, an availability percent from downtime over a period, and the series or parallel availability of dependent services.
Example: 99.9 % allows 43 m 48 s a month and 8 h 45 m 36 s a year; two 99.95 % services in series give 99.9 %, in parallel 99.999975 %.
Nines,
in minutes.
How a percentage becomes minutes of downtime, how a measured outage becomes a percentage, how dependent services combine, and what a contract adds that arithmetic does not.
Availability to downtime
Downtime allowed = (1 − availability) × period. A month here is 365 ÷ 12 days (730 hours) and a year 365 days, so twelve months equal one year exactly. 99.9 % (“three nines”) is 43 m 48 s a month; each extra nine divides the allowance by ten. The reverse question divides the outage you measured by the period you measured it over.
Chains of components
A service that needs every one of its parts (app, database, network, DNS) has the product of their availabilities — always less than the weakest link, and the page’s series mode. Redundant parts where any one suffices have 1 − Π(1 − Aᵢ): two independent 99.95 % nodes in parallel reach 99.999975 %. Parallel assumes independent failures and instant failover; shared power, network or region breaks that assumption.
Contracts
An SLA defines its own measurement period, exclusions (planned maintenance, force majeure), what counts as “down” (error rate, latency thresholds) and the credit schedule. Those terms govern; this page is the arithmetic they rely on, not a reading of any agreement.
What the number does not settle
Mean time between failures and mean time to repair are not inputs — the page takes availability as given. RAID Capacity and Backup Storage cover the storage side of resilience. Inputs stay in your browser; the same four anonymous usage counts as the rest of the site apply.