SLA Uptime Calculator

SLA uptime calculator

Here is the short version: 99.9% sounds like almost always. It allows 43 minutes 12 seconds of downtime in a 30-day month. This SLA uptime calculator turns any percent into day, week, month, and year allowances, then checks how much of this month’s budget you already spent.

It is arithmetic. It is not a vendor credit, an insurance clause, or a promise we will hit five nines on a single firewall. If you want dollars, open the IT downtime cost calculator next.

The formula

Allowed downtime = period length x (1 - uptime / 100)

This page uses:

• Day = 24 hours
• Week = 7 days
• Month = 30 days (2,592,000 seconds)
• Year = 365 days

So 99.9% for a month is 2,592,000 x 0.001 = 2,592 seconds = 43 minutes 12 seconds. A 365-day year is 8 hours 45 minutes 36 seconds. Sites that use 30.44-day months print 43 minutes 50 seconds. Both are right if you label the convention. We label 30 days so the table matches the one in most SLA PDFs. Day and week values round to the nearest second (86.4 seconds becomes 1 minute 26 seconds). A 100% target has a zero-minute budget. Any outage that month is already over it.

Reverse: achieved uptime = 1 - (downtime / period). Fifty minutes down in a 30-day month is 1 - 50/43,200 = 99.884%. That misses 99.9% even though “it was only one incident.”

Google’s SRE book treats the leftover as an error budget. You spend it on deploys and risk, or an incident spends it for you. A firm that never measures the spend is guessing.

Worked example

Target 99.9%. Actual downtime this month: 50 minutes.

• Monthly budget: 43 minutes 12 seconds
• Spent: 50 minutes (about 116% of budget)
• Achieved: 99.884%

One four-hour Microsoft 365 or circuit outage is 240 minutes. That is more than five months of 99.9% budget in a single morning. The percent on the vendor card did not fail. The path to the vendor did.

Microsoft publishes service health and continuity for Microsoft 365, including uptime history and how unplanned incidents are declared. Credits live in the online-services SLA. They do not pay you for a dead local firewall, an expired certificate, or a DNS change that never replicated. The calculator on this page is the office clock, not the tenant clock. Google also publishes a compact availability table of nines if you want the same arithmetic in their units.

What the extra nine actually costs

Do not put 99.99% in a proposal if change windows are “whenever Joe is in.” The number will be used against you. Put 99.9% and then build for it: two power paths where the floor is dark without them, a restore you have timed, and someone who gets the alert at 2 a.m. That is managed infrastructure and backup and disaster recovery, not a slogan.

Copying a four-nines cloud SLA onto a 20-person office with one ISP is how a sales deck becomes a fight after the first outage. Size the pipe with the office bandwidth calculator so the “outage” is not just a saturated upload during backup.

Error budgets make the politics smaller. If the monthly 99.9% budget is 43 minutes, a 15-minute change window is a choice you can still afford. A 90-minute “quick” cutover is not. Write the window on a calendar before anyone opens a maintenance ticket. If the last three incidents were the same host, stop spending the budget on hope and spend it on a second host.

Southern California professional firms feel downtime as missed billable hours and missed appointments. A 43-minute budget disappears in one bad morning of a dark practice-management system. Track minutes even when nobody sends a nasty email. The quiet outages are the ones that train you to think 99.9% is free.

Planned windows still count

Unless the contract excludes announced maintenance, a Saturday cutover spends the same minutes as a crash. If you need the hours, write the exclusion. If you cannot write the exclusion, do not take the whole budget on a change with no rollback.

Track the last four incidents: date, minutes, cause. If three of them are the same firewall or the same host, the SLA conversation is a purchasing conversation.

Who owns the clock

A vendor SLA counts their service. Your users count “could I work.” If Exchange Online is green and the office firewall is looping, the tenant SLA is intact and the firm is still down. Measure the path the attorney actually uses: identity, DNS, the circuit, the Wi-Fi, then the app. The office bandwidth calculator is the pipe. This page is the minutes. The downtime cost calculator is the money.

Keep a running total in a shared sheet: date, minutes, what broke, who noticed. After a quarter you will know whether 99.9% is a slogan or a description. If every row is “ISP” you have a circuit problem. If every row is “host” you have an infrastructure problem. If every row is “nobody knew for two hours” you have a monitoring problem. The percent does not tell you which one. The list does.

Do not average minutes across a year to hide a bad month. SLAs are usually monthly. A quiet November does not buy you a four-hour December.

What this calculator is not

It is not a credit calculator. Vendors prorate, cap, and exclude. It is not five nines on a single ISP. It is not a substitute for logging actual downtime. Memory rounds down. Tickets do not.

Use it to translate a percent into a budget a non-engineer can feel. Then decide whether last month’s 50 minutes was a fluke or the process.

If the budget is already gone and the next close is Monday, contact Secure Techies. Bring the outage list. Minutes first. Dollars second.