We promised 99.9% uptime — how much downtime is that, and are we actually meeting it?
Q: We signed an SLA at three nines but nobody calculated what it allows. How do I check if we're compliant?
A: Three nines is 43 minutes 12 seconds of allowed downtime per 30-day month, or about 8 hours 46 minutes per year. People underestimate how tight that is.
A reader audited their own numbers and got a surprise. Their dashboard claimed 99.95%, but it only counted full outages. When they added partial degradations — checkout slow enough to time out for 5% of users — their effective availability for paying customers was 99.82%. That breached the SLA and they hadn't noticed because the math excluded brownouts.
They rebuilt the calculation to weight degraded periods, and used the real 99.82% to justify an infrastructure fix that pushed them back to 99.94%.
The follow-up: define in writing what counts as 'down.' Most SLA disputes are really definition disputes.
Got a question? Drop it in the comments.
Pingback Clinic
@PingbackClinic
We promised 99.9% uptime — how much downtime is that, and are we actually meeting it?
Этот пост опубликован в Telegram-канале Pingback Clinic. Подписаться можно по ссылке: @PingbackClinic.