← All calculators
Availability and SLO calculator
Combine components in series and in parallel, get overall availability, downtime per month and year, error budget minutes and SLO burn rate.
Components
Every block is in series: a request needs all of them. Inside a redundant group, any one member is enough.
99.99%
99.9999875%
Member 1
Member 2
Member 3
99.9999%
Member 1
Member 2
99.95%
Combined availability
- System availability
- 99.9399%3.2 nines
- Downtime per month
- 26 min30-day month
- Downtime per year
- 5.27 h
Per block (downtime per year)
- Load balancer99.99% · 52.6 min
- App servers99.9999875% · 3.94 s
- Database primary + standby99.9999% · 31.5 s
- Payment provider weakest99.95% · 4.38 h
Error budget and burn rate
Failed ÷ total requests right now
- Error budget
- 43.2 min0.1% of 30 days
- Architecture uses
- 60.1%26 min expected downtime
- Burn rate
- 5×1× spends the budget exactly over the window
- Budget gone in
- 6 daysat this error rate
- Spent in 1 h
- 0.694%of the whole window’s budget
How this is calculated
- Series
A = A₁ × A₂ × … × Aₙ. Every dependency you add in the request path lowers the total.- Parallel (redundant)
A = 1 − (1 − A₁) × (1 − A₂) × …. This assumes failures are independent and failover is instant; shared dependencies, bad deploys and slow failover break that assumption.- Downtime
downtime = (1 − A) × period, with a 30-day month (43,200 min) and a 365-day year.- Error budget and burn rate
budget = (1 − SLO) × window,burn rate = error rate ÷ (1 − SLO),time to exhaust = window ÷ burn rate. A common paging rule is a burn rate around 14 over an hour (2% of a 30-day budget gone) or around 6 over six hours.
What to say in the interview
- “Chaining these in series gives about 99.9399%, which is 26 min of downtime a month; the weakest block is Payment provider, so that’s where redundancy pays off most.”
- “Redundancy only multiplies out like this if the replicas fail independently. In practice they share deploys, config and a region, so I’d treat the parallel numbers as an upper bound.”
- “With a 99.9% SLO we get 43.2 min of error budget per 30 days. A 0.5% error rate is a 5× burn, which empties the budget in 6 days, so that should page someone.”