99.99% uptime — four nines.
99.99% uptime means less than 53 minutes of downtime per year. Under 9 seconds per day. This is where things get serious. A single bad deployment that takes 10 minutes to roll back blows through your monthly budget. You can't achieve four nines by being good at fixing things. You need to be good at not breaking things.
Maximum allowed downtime
Per day
8.6s
Per week
1m 0.5s
Per month
4m 21s
Per year
52m 33s
WHO NEEDS IT
Who targets 99.99%
Payment processing, authentication services, core APIs that other services depend on, e-commerce platforms during peak seasons, anything where minutes of downtime translate to real revenue loss.
WHAT IT TAKES
How to achieve four nines
Load balancing across multiple availability zones, automated failover, zero-downtime deployments, comprehensive monitoring with sub-minute detection, on-call rotation, and an engineering culture that prioritizes reliability. Most outages at this level come from deployments, not hardware.
Other uptime tiers
Measure your actual uptime.
Larm uses multi-probe voting to verify outages before alerting — so false positives don't eat into your uptime budget.