−50%

50% off your first month on any paid plan · code WELCOME50 · until Nov 30

Start free →

Free tool · Calculator

Uptime Calculator

Turn an uptime percentage into allowed downtime per day, week, month and year.

Calculator free · no sign-up

Common targets

Per day
—
Per week
—
Per month
—
Per quarter
—
Per year
—

Time period

Uptime %
—
Nines level
—
Budget left/month
—

What do the nines mean in uptime SLAs?

This uptime calculator (also works as an SLA calculator and SLA downtime calculator) converts any uptime percentage into allowed downtime per day, week, month, and year. The gap between what a contract says and what it means in minutes is where most SLA disputes start.

The difference between 99.9% and 99.99% uptime looks small, just 0.09 percentage points. In minutes, that's 43 minutes of downtime per month versus four and a half. One extra decimal digit divides the allowed downtime by ten. Tenfold. That's the reason SLA commitments use percentages rather than vague terms like "high availability."

How much downtime is 99.9% uptime?

99.9% uptime (three nines) allows 8 hours and 45 minutes of downtime per year. About 43 minutes per month, or 1 minute 26 seconds per day. It's the most common SLA tier for cloud services, managed hosting, and SaaS platforms. Most teams can realistically hit three nines with standard infrastructure: a single load balancer, health checks, automated restarts. No redundant data centers required.

Where it gets tight: 43 minutes per month includes everything. Planned deployments, database migrations, the five minutes your config was wrong before anyone noticed. We track uptime for thousands of monitors on WatchCron, and the services that dip below three nines usually share one culprit: deployments without health checks. If you deploy twice a week and each deploy causes 30 seconds of downtime, that's already 4 minutes gone before anything actually breaks.

How much downtime is 99.99% uptime?

Zero-downtime deployments stop being optional at four nines. 99.99% uptime allows 52 minutes of downtime per year, which works out to 4 minutes and 23 seconds per month. You need blue-green or rolling deployments, automated failover, and an on-call rotation that can respond in under two minutes. That's a real operational cost most teams underestimate until they're contractually bound to it.

Four nines is what enterprise customers expect when they're paying five or six figures annually. It's also where SLA credits get expensive. A single bad deploy that takes your service down for 15 minutes? That's 3.4x the monthly allowance, gone in one shot. At a typical 10% credit per 0.1% breach, that's a painful line item when multiple customers file claims in the same billing cycle.

How much downtime is 99.999% uptime?

Very few services actually deliver five nines in practice. The number itself sounds clean: 5 minutes and 15 seconds of downtime per year. Twenty-six seconds per month. At this level, even a slow health check response might count as an outage.

Five nines demands multi-region active-active infrastructure, failover that detects problems within milliseconds, and a team that treats every hiccup as a potential incident. AWS S3 promises 99.99% (four nines, not five). Google Cloud Spanner advertises 99.999%, but that's the database layer alone; your application on top of it will have its own failure modes.

If someone offers you five nines for $50/month, they're either lying or defining "downtime" creatively.

99.9% vs 99.95% uptime: what's the real difference?

When you're choosing between hosting providers, one promises 99.9%, the other 99.95%. The calculator translates that to: 43 minutes per month versus 22 minutes. If your service handles payments, those 20 minutes are the difference between an acceptable incident and chargebacks piling up.

Before promising customers 99.99% in your own SLA, it's worth seeing that number in minutes: four and a half per month. That's including updates, restarts, DNS hiccups. We've seen teams promise four nines in contracts written before they even had automated failover. The first real incident usually triggers a painful renegotiation.

Real SLA commitments from major cloud providers

What the big providers actually promise (not what they achieve, what they're contractually liable for):

  • AWS EC2: 99.99% per region (52 min/year). Credits range from 10% for 99.0–99.99% down to 30% below 95%. By most third-party estimates, actual delivery hovers around 99.995%.
  • Google Cloud Compute Engine: 99.99% per region. Credits scale from 10% to 50% depending on actual uptime.
  • Azure Virtual Machines: 99.95% for Availability Sets (4h 23m/year), 99.99% for Availability Zones.
  • Vercel: 99.99% for Enterprise. No SLA at all on free or Pro plans.
  • Cloudflare: 100% for Enterprise (they refund 25x the downtime as credits). Business plan sits at 99.99%.
  • Heroku: 99.95% for Production/Enterprise dynos. Free and hobby? Nothing.

Notice the pattern. The SLA commitment is always lower than actual uptime. The gap protects the provider. When evaluating an SLA, the useful question isn't "what do they promise?" It's "what happens when they miss it, and does the credit clause actually cover my losses?"

SLA downtime table: 99% to 99.999%

The table below shows how much downtime each SLA level allows per day, week, month, and year, assuming a 24/7 monitoring window.

SLA levelDowntime/dayDowntime/weekDowntime/monthDowntime/year
99% (two nines)14m 24s1h 41m7h 18m3d 15h 36m
99.5%7m 12s50m 24s3h 39m1d 19h 48m
99.9% (three nines)1m 26s10m 5s43m 50s8h 45m 36s
99.95%43s5m 2s21m 55s4h 22m 48s
99.99% (four nines)8.6s1m 0.5s4m 23s52m 34s
99.999% (five nines)0.86s6s26s5m 15s

If your SLA excludes planned maintenance or off-hours, the effective allowed downtime is higher. But so is the expectation that covered hours are truly uninterrupted. Some providers play this game well, carving out "scheduled maintenance windows" that conveniently don't count against their published number.

How to calculate uptime percentage from downtime

The formula: Uptime % = ((Total minutes − Downtime minutes) / Total minutes) × 100.

A 30-day month has 43,200 minutes. If you had 45 minutes of downtime, your uptime was 99.896%. Just below three nines. One bad incident and the whole month falls short. For spreadsheets: =((43200-B2)/43200)*100 where B2 is your downtime in minutes.

Got a specific number of downtime minutes and want to know your actual uptime percentage? The reverse uptime calculator above does that automatically. Enter the downtime, pick the period, and it converts back. Useful for post-incident reviews and SLO compliance checks.

Nobody does this by hand anymore. Uptime monitoring logs every check automatically, and uptime reports turn that data into an exact percentage for any time range. The reports are shareable, which matters when someone in a quarterly review asks "so what was our actual uptime?" To put a dollar figure on that downtime, the downtime cost calculator runs those numbers.

What happens when you breach your SLA

Breach the number and credits kick in. If you promise 99.9% and deliver 99.8%, customers can claim a credit, typically 10–30% of their monthly bill per 0.1% below the target. For a $10K/month contract, a 99.5% month could mean $4K–$12K in credits. That adds up fast when multiple customers file claims in the same cycle.

The financial hit from credits is often the smaller cost. Churn, lost trust, emergency engineering hours. Those don't show up on the invoice but they compound. Honestly, three nines is the right target for most teams. If you're promising four nines before you have automated failover and a tested runbook, you're writing checks your infrastructure can't cash.

An SLA isn't a promise. It's the penalty clause for breaking one. Pick the number you can actually hit, monitor it with real uptime data, and keep an incident log that proves you took every breach seriously. Our uptime reports round to two decimal places; for five-nines tracking you'd need sub-second granularity that most monitoring tools, ours included, don't yet provide out of the box. That's worth knowing before you sign the contract.

Frequently asked questions

99.9% uptime (three nines) allows 8 hours 45 minutes of downtime per year, or about 43 minutes per month. This is the most common SLA tier for cloud hosting and SaaS. Most teams can hit it with standard infrastructure.

99.99% uptime (four nines) allows 52 minutes of downtime per year, just 4 minutes 23 seconds per month. Reaching this level requires zero-downtime deployments, automated failover, and an on-call rotation.

Five nines means 99.999% uptime: only 5 minutes 15 seconds of total downtime per year, or 26 seconds per month. Very few services achieve this. It requires multi-region redundancy and sub-second failover detection.

99.95% uptime allows 4 hours 22 minutes of downtime per year, or about 22 minutes per month. This is the SLA level Azure offers for Availability Sets and Heroku offers for Production dynos.

99% uptime (two nines) allows 3 days 15 hours of downtime per year. That is over 7 hours per month. Common for free-tier hosting and hobby projects where some downtime is acceptable.

Uptime % = ((total minutes − downtime minutes) / total minutes) × 100. For a 30-day month with 45 minutes of downtime: (43,200 − 45) / 43,200 × 100 = 99.896%. The reverse calculator above does this automatically.

99.9% (three nines) is realistic for most small teams. It allows about 43 minutes of downtime per month. Promising 99.99% without redundant systems and on-call rotation is risky.

An SLA calculator (also called an uptime calculator) converts an uptime percentage into concrete downtime allowances per day, week, month, and year. It shows what a service-level agreement means in actual minutes.

More free tools

All 20 tools →

Turn a one-off check into a monitor.

feature
Pricing Docs
Log in Start free →