If you have ever worked on a project that uses components based in the cloud, you would remember coming across the term uptime. This is often along the lines of 99.9% or 99.99% or 99.999% or anything else that looks ridiculously close to 100% in the layman's eyes.
If you don't work in the cloud, then you would mostly think how tiny a difference exists between 99.9% and 99.99%. However, that is not exactly the case. The difference between each of these values are pretty large and can often lead to make / break situations for companies using the concerned cloud product.
### Firstly, these metrics are Service Level Agreements (SLAs)
An SLA is a *plain language agreement* between a vendor (cloud provider) and a customer (businesses or individuals). This agreement defines
- What services will be delivered
- Responsiveness to queries / issues
- Performance measurement
Putting it simply, an SLA is a metric that is inversely proportional to the *perceived IT headache a business might need to go through if moving into the cloud*. Higher the SLA value, lower the headache. And, businesses love avoiding IT headaches altogether.
However, there is always that rare time when things break. Hence an SLA is almost never 100%. But, the *closest it is to 100%*, the happier customers are. Vendors also need to ensure that these SLAs are met. If they fail to do so, they might be liable to compensate their customers. And, vendors love avoiding such glitches altogether.
### Now, to the quantified meaning of an uptime percentage
An uptime percentage is a measure of reliability of a system - the percentage of time a system is ready for operation under normal circumstances.
$
\text{Uptime Percentage} = \left( \frac{\text{Total Time} - \text{Downtime}}{\text{Total Time}} \right) \times 100
$
**Example**
If a vendor suggests a system has 99.9% uptime, then let's calculate how much downtime is expected per month.
$
99.9 = \left( \frac{2592000 - \text{Downtime}}{2592000} \right) \times 100
$
$
\text{Downtime} = \text{43 minutes 28 secs per month}
$
Let's do the same for a few other cases. If you, like me are lazy to do math, just head over to [uptime.is](https://uptime.is/) for computing uptime.
| Agreed SLA Level | Downtime |
| ---------------- | ---------------------------------------------------------------------------------------------------------------------------------------------- |
| 99% | - **Daily:** 14m 24s<br>- **Weekly:** 1h 40m 48s<br>- **Monthly:** 7h 14m 41s<br>- **Quarterly:** 21h 44m 4.4s<br>- **Yearly:** 3d 14h 56m 18s |
| 99.9% | - **Daily:** 1m 26s<br>- **Weekly:** 10m 4.8s<br>- **Monthly:** 43m 28s<br>- **Quarterly:** 2h 10m 24s<br>- **Yearly:** 8h 41m 38s |
| 99.99% | - **Daily:** 8.6s<br>- **Weekly:** 1m 0.48s<br>- **Monthly:** 4m 21s<br>- **Quarterly:** 13m 2.4s<br>- **Yearly:** 52m 9.8s |
| 99.999% | - **Daily:** 0.86s<br>- **Weekly:** 6s<br>- **Monthly:** 26s<br>- **Quarterly:** 1m 18s<br>- **Yearly:** 5m 13s |
While 99.9% and 99.999% look quite close in the layman's eyes, these in fact are quite drastically different.