Uptime SLA Calculator

What the nines mean in real time

99 percent allows 3 days 15 hours a year. 99.9 percent, the common commercial default, allows 8 hours 45 minutes a year, or about 43 minutes a month. 99.95 allows 4 hours 22 minutes annually. 99.99 allows 52 minutes a year, roughly 4 minutes 20 seconds a month. 99.999 allows 5 minutes 15 seconds a year — less time than it takes a human to read an alert, diagnose and act, which is why five nines implies automated failover rather than a responsive team.

Each nine costs roughly an order of magnitude

Going from 99.9 to 99.99 means removing 90 percent of remaining downtime, and the cheap causes have already been fixed. It typically requires multi-region redundancy, automated failover, zero-downtime deployment and elimination of every single point of failure including the database. Most organisations discover that the target they can afford is a nine below the one they promised.

The measurement window is the negotiation

99.9 percent measured monthly allows 43 minutes each month. Measured annually it allows 8 hours 45 minutes, which can all fall in one incident. Annual measurement is much weaker for the customer, and is what most providers offer. What counts as an outage matters just as much: partial degradation, elevated error rates and slow responses are frequently excluded, so a service can be unusable and technically compliant.

Dependencies multiply, they do not average

A service depending on three components each at 99.9 percent has a theoretical maximum of 99.7 percent, because any one failing takes it down. Adding dependencies lowers the ceiling, which is why an architecture with many synchronous downstream calls cannot exceed the weakest link. Redundancy inverts this — two independent components at 99 percent in parallel give 99.99 percent, provided the failures really are independent, which shared infrastructure quietly prevents.

Maintenance windows and exclusions

Most SLAs exclude scheduled maintenance, force majeure, customer-caused issues and problems in third-party networks. Read those clauses before comparing headline percentages: an SLA with generous exclusions and a higher number can deliver worse availability than a stricter one with a lower number.

The remedy is usually trivial

SLA breaches typically pay out as service credits worth a small percentage of the monthly fee, capped well below the customer's actual loss. The SLA is a statement of intent and a pricing signal, not insurance. If downtime would cost you materially more than the credit, the mitigation has to be your own architecture rather than the contract.