99.95 uptime calculator
At 99.95% uptime the calculator gives 43.2s of allowed downtime per day, 5m 2.4s per week, 21m 54.9s per month, 1h 5m 44.7s per quarter and 4h 22m 58.8s per year.
99.95% uptime allows 4h 22m 58.8s of downtime a year.
| Period | Downtime allowed |
|---|---|
| Day | 43.2s |
| Week | 5m 2.4s |
| Month | 21m 54.9s |
| Quarter | 1h 5m 44.7s |
| Year | 4h 22m 58.8s |
99.95 is the number cloud providers like. Google Cloud Run promises it for non-GPU services in most regions, measured per calendar month. It sits between three and four nines and allows exactly half the downtime of 99.9. The calculator assumes 24/7 service, a 365.25-day year and a 30.44-day month.
99.95 uptime: the full table
- Per day: 43.2 seconds
- Per week: 5m 2.4s (302.4 seconds)
- Per month: 21m 54.9s (1,314.9 seconds)
- Per quarter: 1h 5m 44.7s (3,944.7 seconds)
- Per year: 4h 22m 58.8s (15,778.8 seconds)
99.95 uptime per month
21m 54.9s on an average month. That is one slow rollback, or one database failover that takes longer than the docs promised. Per year the same target allows 4h 22m 58.8s, but a yearly reading lets one long outage hide inside eleven clean months, so most SLAs at this level reset monthly.
99.95 SLA downtime per calendar month
# 99.95% per calendar month, the way cloud SLAs measure it.
from calendar import monthrange
target = 99.95
for month in (2, 4, 1): # 28, 30 and 31 days in 2026
days = monthrange(2026, month)[1]
allowed = days * 86400 * (1 - target / 100)
print(f"{days} days: {allowed // 60:.0f}m {allowed % 60:.1f}s")
# 28 days: 20m 9.6s
# 30 days: 21m 36.0s
# 31 days: 22m 19.2sCloud SLAs count calendar months, so the real allowance moves between 20m 9.6s in February and 22m 19.2s in a 31-day month. Cloud Run credits 10% of the monthly bill when a month lands between 99% and 99.95%, 25% between 95% and 99%, and 50% below 95%, and states that the credit is your only remedy. Your customers get none of it, and that outage still counts in full against your own SLA.
Your 99.95 sits on top of theirs
If your app runs on a 99.95 host and talks to a 99.95 database, the best you can promise without redundancy is 99.95% times 99.95%, which is 99.90%. Two dependencies at the same target halve your budget before you write any code. Promise 99.9 on top of a 99.95 platform, not 99.95.
Measuring 99.95 with checks
Logdash counts successful checks over all checks, a 200 to 399 status within 10 seconds being a success. At 5 minutes the month holds 8,766 checks and 4 may fail. At 1 minute, on Builder, 21 may fail. At 15 seconds, on Pro, 87 may fail. The 5-minute interval is too coarse to trust here: a 6-minute outage shows up as one or two failed checks, a quarter or half of the monthly budget, depending on where the checks happened to land.
- Point a monitor at the app Create a service in Logdash and give it your public URL or a health route. The first check runs immediately and records the status code and response time.
- Pick 1 minute or faster Builder checks every minute for $9 a month and Pro every 15 seconds for $15. The free plan checks every 5 minutes, enough to see 99.9 but not to trust a 99.95 number.
- Get told inside the budget Connect a Telegram channel and stop the app. The next check fails, the monitor flips to down, and Telegram delivers the alert with the monitor name and the status code while almost all of the 21 minutes is still yours.
Measuring 99.95: Logdash vs Hyperping
| Feature | Logdash | Hyperping |
|---|---|---|
| Free plan | 5 services, checked every 5 minutes | 20 monitors, checked every 5 minutes |
| Fastest paid interval | 15 seconds on Pro | 30 seconds on Essentials, 20 seconds on Business |
| Failed checks inside 21m 55s | 87 at 15 seconds | 43 at 30 seconds |
| Status page on the free plan | One public status page | One basic status page |
When Hyperping is the better pick
- You have more than five things to watch and no budget yet. The Hyperping free plan covers 20 monitors at the same 5-minute interval.
- You want monitors, status pages and incident management sold as one bundle, which is how Hyperping is built.
- 30 seconds is enough for your 99.95. Forty-three failed checks a month is still a fine-grained picture.