99% uptime: how much downtime is that?
99% uptime allows 14 minutes 24 seconds of downtime a day, 7 hours 18 minutes in an average month and 3 days 15 hours 36 minutes a year. It is the loosest target most teams write down, and one that problems fixed during working hours can meet.
Downtime allowed by 99% uptime two nines
| Period | Allowed downtime |
|---|---|
| Daily | 14m 24s |
| Weekly | 1h 40m 48s |
| Monthly | 7h 18m |
| Quarterly | 21h 54m |
| Yearly | 3d 15h 36m |
| 30-day month | 7h 12m |
| 31-day month | 7h 26m 24s |
| 28-day month | 6h 43m 12s |
A month is the average month, a twelfth of a 365-day year; a quarter is a quarter of it.
What that looks like
A whole night of downtime fits in a month: if the site goes down at midnight and someone fixes it at 7 a.m., the month is still within 99%.
Measured per day the budget is much tighter. A day with a 15-minute outage already misses a daily 99%, so check which period your target uses.
How many outages fit
| One outage of | Per month | Per year |
|---|---|---|
| 1m | 438 | 5256 |
| 5m | 87 | 1051 |
| 15m | 29 | 350 |
| 30m | 14 | 175 |
| 1h | 7 | 87 |
| 4h | 1 | 21 |
Counts whole outages of that length that fit in the budget of an average month or a year, with nothing else going wrong.
What 99% takes
- One server or one region is enough; nobody needs to be on call.
- Alerts by email or chat that someone reads at the start of the next working day.
- A process manager or container restart policy, so a crash restarts on its own.
- Backups you have restored at least once, because a full rebuild can take most of a month's budget.
Where it fits: Internal tools, staging environments, side projects and batch jobs where a delay of a few hours costs little.
How to measure it
Check at most every tenth of the budget, so short outages do not slip between checks: every 43m 48s or more often for a monthly 99%. In OpenUptime that is any plan, including Free (every 5 minutes).
Check from more than one region, so a single network problem between one location and your server is not counted as downtime.
Compared with other targets
| Uptime | Monthly | Yearly | Downtime against 99% |
|---|---|---|---|
| 99% | 7h 18m | 3d 15h 36m | this page |
| 99.5% | 3h 39m | 1d 19h 48m | 1/2 as much |
| 99.9% | 43m 48s | 8h 45m 36s | 1/10 as much |
| 99.95% | 21m 54s | 4h 22m 48s | 1/20 as much |
| 99.99% | 4m 23s | 52m 34s | 1/100 as much |
| 99.999% | 26.3s | 5m 15s | 1/1000 as much |
Track 99% uptime
OpenUptime records every check, shows uptime per monitor and on 90-day status page bars, and alerts you when a check fails.
Questions
Is 99% uptime good?
For an internal tool, yes. For a public product it is low: 7 hours a month is long enough for customers to notice and to ask about it.
What is the difference between 99% and 99.9%?
Ten times the downtime. 99% allows 3 days 15 hours a year, 99.9% allows 8 hours 45 minutes.
Can a single server reach 99%?
Often, if restarts and updates stay short and someone notices a failure within hours. A disk or host failure that needs a rebuild from backup is the usual way a single server misses it.
99% 可用率意味着多少停机时间？
99% 可用率每天允许 14 分 24 秒停机，平均每月 7 小时 18 分，每年 3 天 15 小时 36 分。这是大多数团队会写下的最宽松的目标，在工作时间内处理问题就能达到。
99% 可用率允许的停机时间 2 个 9
| 周期 | 允许的停机时间 |
|---|---|
| 每天 | 14 分 24 秒 |
| 每周 | 1 小时 40 分 48 秒 |
| 每月 | 7 小时 18 分 |
| 每季度 | 21 小时 54 分 |
| 每年 | 3 天 15 小时 36 分 |
| 30 天 | 7 小时 12 分 |
| 31 天 | 7 小时 26 分 24 秒 |
| 28 天 | 6 小时 43 分 12 秒 |
一个月指平均月，即 365 天的十二分之一；一个季度是一年的四分之一。
实际意味着什么
一个月里可以容纳一整夜的停机：如果网站半夜 12 点宕机，早上 7 点有人修好，这个月仍然在 99% 以内。
按天计算时，预算就紧得多。某一天出现 15 分钟故障，就已经达不到按天计算的 99%，所以要先确认目标是按哪个周期计算的。
预算能容纳几次故障
| 单次故障时长 | 每月 | 每年 |
|---|---|---|
| 1 分 | 438 | 5256 |
| 5 分 | 87 | 1051 |
| 15 分 | 29 | 350 |
| 30 分 | 14 | 175 |
| 1 小时 | 7 | 87 |
| 4 小时 | 1 | 21 |
表示在平均月或一年的预算内，能容纳多少次这样时长的完整故障（假设没有其他问题）。
做到 99% 需要什么
- 一台服务器或一个区域就够了，不需要有人值班。
- 通过邮件或聊天工具发出的告警，有人在下一个工作日开始时查看即可。
- 使用进程管理器或容器重启策略，程序崩溃时能自动重启。
- 至少实际恢复过一次的备份，因为一次完整重建可能用掉一个月的大部分预算。
适用场景： 内部工具、预发布环境、个人项目，以及延迟几个小时影响不大的批处理任务。
怎样测量
检查间隔不要超过预算的十分之一，以免短暂的故障落在两次检查之间：按月计算的 99%，检查间隔应不超过 43 分 48 秒。在 OpenUptime 中，这对应 任何套餐，包括 Free（每 5 分钟）。
从多个区域检查，这样某一个地点到服务器之间的网络问题就不会被算作停机。
与其他目标对比
| 可用率 | 每月 | 每年 | 停机时间与 99% 相比 |
|---|---|---|---|
| 99% | 7 小时 18 分 | 3 天 15 小时 36 分 | 本页 |
| 99.5% | 3 小时 39 分 | 1 天 19 小时 48 分 | 1/2 |
| 99.9% | 43 分 48 秒 | 8 小时 45 分 36 秒 | 1/10 |
| 99.95% | 21 分 54 秒 | 4 小时 22 分 48 秒 | 1/20 |
| 99.99% | 4 分 23 秒 | 52 分 34 秒 | 1/100 |
| 99.999% | 26.3 秒 | 5 分 15 秒 | 1/1000 |
常见问题
99% 的可用率算好吗？
对内部工具来说算好。对面向公众的产品来说偏低：每月 7 小时足以让客户注意到并来询问。
99% 和 99.9% 差多少？
停机时间相差十倍。99% 每年允许 3 天 15 小时，99.9% 每年允许 8 小时 45 分。
一台服务器能达到 99% 吗？
通常可以，只要重启和更新都很快，故障能在几小时内被发现。单台服务器达不到 99%，最常见的原因是磁盘或主机故障，需要从备份重建。