When your hosting provider promises "99.9% uptime" in their SLA, that sounds impressive. Three nines. Nearly perfect. But what does it actually mean in practice — and should you accept it?
The answer depends entirely on what you're building. For a personal portfolio, 99.9% is fine. For an e-commerce checkout or a SaaS product with paying customers, it might not be nearly good enough.
Here's everything you need to know.
Uptime percentages are calculated against a full year (8,760 hours). The remaining percentage becomes your maximum allowed downtime.
| Uptime % | Called | Downtime / year | Downtime / month | Downtime / week |
|---|---|---|---|---|
| 99% | Two nines | 87.6 hours | 7.3 hours | 1.68 hours |
| 99.9% | Three nines | 8.76 hours | 43.8 minutes | 10.1 minutes |
| 99.95% | Three and a half | 4.38 hours | 21.9 minutes | 5 minutes |
| 99.99% | Four nines | 52.6 minutes | 4.4 minutes | 1 minute |
| 99.999% | Five nines | 5.26 minutes | 26 seconds | 6 seconds |
Let that sink in for a moment. A provider with a 99.9% SLA is contractually allowed to be down for almost 9 hours per year. That could be a single 9-hour outage or dozens of smaller ones scattered throughout the year — and they'd still be within their SLA.
Most shared hosting providers, cheaper cloud tiers, and many managed services promise 99.9%. Here's what that actually looks like day-to-day:
For many workloads, this is acceptable. But for any site where downtime = lost revenue, three nines is a floor, not a target.
Quick check: If your site was down for 44 minutes during business hours, what would it cost you? If the answer is "a lot," you need a higher tier — and more importantly, you need to verify your provider is actually hitting their SLA.
99.99% uptime means just 52 minutes of downtime per year — less than a minute per week. This is the standard for most production SaaS products, payment processors, and anything customer-facing with a direct revenue impact.
Getting from three nines to four nines isn't just a configuration change. It typically requires:
This is also why four-nine infrastructure costs significantly more. You're paying for the redundancy, not just the compute.
99.999% — five nines — means 5 minutes of downtime per year. This is the standard for telephone networks, financial trading systems, and critical healthcare infrastructure. At this level, even planned maintenance windows are a problem: you can't take anything offline for updates without a hot-standby replacement ready.
Five nines is expensive, complex, and genuinely hard to achieve. Most SaaS products don't need it, and the engineering cost to get there often isn't justified unless downtime has life-or-safety implications or massive direct financial consequences.
Here's the thing about SLAs: they're a legal contract for compensation, not a guarantee of actual behavior. Your provider can hit their 99.9% SLA and still have outages that cause real damage to your business — the SLA just determines whether you get a service credit.
More importantly, SLAs typically exclude:
A provider can be down for 2 hours during a "scheduled maintenance window" and that doesn't count against their 99.9% SLA at all. You need to measure your actual uptime from the outside — from your users' perspective — not just trust your provider's dashboard.
This is exactly why uptime monitoring exists. Your provider's internal metrics don't capture DNS failures, CDN edge issues, or partial outages that only affect certain regions. External monitoring does.
Here's a practical framework based on what you're building:
99.9% is fine. A few hours of downtime per year won't meaningfully impact anything. Use a cheap hosting tier and don't lose sleep over it.
Target 99.9% in practice, with a plan to get to 99.95% as you grow. The investment in redundancy at this stage is premature — focus on product instead. But do set up monitoring so you know when you're down.
You need 99.99% as your target and should be actively measuring against it. At this point, an hour of downtime has a real cost — in refunds, churn, and support load — that justifies spending on better infrastructure.
99.99% is the baseline expectation. Many enterprise procurement teams won't sign a contract without it. Work toward 99.999% for core functionality, even if peripheral features can tolerate less.
The only way to know your real uptime is to monitor it from the outside. Internal health checks and provider dashboards don't reflect what your users actually experience.
You need a monitoring tool that:
Once you have that data, calculating your uptime is simple: divide the number of successful checks by total checks over any time period. PingSentry shows this automatically across 1h, 24h, 7d, and 30d windows on every monitor's detail page.
Uptime SLA percentages matter — but they're a ceiling on liability, not a guarantee of reliability. The only way to know where you actually stand is to measure it yourself.
Three nines sounds great until you realize it's 9 hours of potential downtime per year. Four nines is the real production standard. Five nines is for systems where failure isn't an option.
Whatever tier you're targeting, the first step is knowing your current baseline — and that requires external monitoring running 24/7.
PingSentry checks your site every 30 seconds and shows your real uptime percentage across 1h, 24h, 7d, and 30d windows. Free tier included.
Start Monitoring Free →