Translate an availability target into the downtime it actually permits — and work out what uptime you hit.
99.9% gives you about 43 minutes a month. An outage that starts at 2am and is noticed at 8am has already spent eight times that. The number only means something if something is watching the clock — from outside your own infrastructure, where a failure cannot take the watcher down with it.
How Pingura measures uptimeRoughly 43 minutes and 49 seconds per month, or about 8 hours 45 minutes per year. That is calculated against an average month of 30.44 days. Many people are surprised that "three nines" permits the better part of a working day of outage annually.
A factor of ten. 99.9% allows about 43 minutes of downtime a month; 99.99% allows about 4 minutes 22 seconds. That difference is usually the line between "someone notices and fixes it" and "it has to fail over automatically".
No, and SLAs that promise it are describing a credit policy rather than an engineering guarantee. Every system has deploys, dependency failures, and certificate renewals. The useful question is how much downtime you can absorb and how quickly you find out it is happening.
It depends on the contract. Most commercial SLAs exclude announced maintenance windows, which is why the advertised number and the availability your users actually experience can differ. If you publish an SLA, define the exclusions explicitly.
You need something checking from outside your own infrastructure at a known interval, recording every failure. Internal health checks cannot report an outage that takes down the thing doing the reporting. Pingura checks from five global regions and keeps the history the calculation needs.
Free plan includes 15 monitors and a public status page. No credit card.
Get Started Free