What 99.9% Uptime Actually Means
SLA percentages compress into a single number that hides just how different they are in practice. Here's what they translate to, and what actually matters when reviewing a contract.
The numbers, laid out
Downtime allowance scales non-linearly as you add nines:
99% ≈ 3.65 days / year
99.9% ≈ 8.76 hours / year
99.95% ≈ 4.38 hours / year
99.99% ≈ 52.6 minutes / year
99.999% ≈ 5.26 minutes / year
Going from 99% to 99.9% is a 10x reduction in allowed downtime. Going from 99.9% to 99.99% is another 10x. The jump from "two nines" to "three nines" looks small on paper (99% vs 99.9%) but represents a completely different operational reality — 3.65 days versus under 9 hours a year.
Why this matters more than it seems
Most cloud IaaS providers commit to 99.9%–99.99% for basic compute instances, and reserve higher tiers (99.95%+) for multi-zone or managed database services. If you're building something that depends on multiple vendor SLAs stacked together — say, a compute SLA at 99.9% plus a database SLA at 99.95% plus a CDN SLA at 99.99% — your actual achievable uptime is bounded by the weakest link, not the average. Stacking three "very good" SLAs doesn't give you a "very very good" result; it gives you roughly the worst one, since any component failing takes the whole system down.
What to actually check in a vendor SLA — beyond the headline number
Does downtime exclude scheduled maintenance? Many contracts define "downtime" as only unplanned outages — scheduled maintenance windows don't count against the SLA at all, even if your service is unavailable during them. Check whether maintenance windows are announced with enough notice to plan around.
What's the measurement window? A 99.9% monthly SLA and a 99.9% annual SLA are not equivalent in practice — a bad week that blows the monthly target resets every month, while an annual target lets a single bad incident be absorbed by the rest of the year.
What's the actual remedy? SLA credits are typically a percentage of monthly fees, tiered by how far below target the actual uptime fell — and they're almost never proportional to the business impact of the outage itself. Read the credit schedule, not just the headline percentage, before treating an SLA as a real guarantee of business continuity.
Need to convert a specific SLA number into concrete downtime, including quarterly and weekly breakdowns? Use the SLA Uptime Calculator.