Availability (also known as service availability) is a metric commonly used to quantitatively measure reliability.
This percentage is calculated in periods such as a month, a year, or the last three years. Applying the strictest possible interpretation, availability decreases whenever the application is not operating normally, including scheduled and unscheduled outages.
The following is a table of common application availability design goals and the maximum duration of outages that can occur in a year without the goal being reached. The table contains examples of common application types at each availability level.
| Availability Objective | Max. Unavailability (by year) | RTO | # of Regions | Application Type |
|---|---|---|---|---|
| 90% | 36 days and 12 hours | 4 - 24 hours | One | . |
| 95% | 18 days and 6 hours | 4 - 12 hours | One | . |
| 99% | 3 days and 15 hours | 2 - 8 hours | One | batch processing tasks, data analysis |
| 99.9% | 8 hours 45 minutes | 1 - 4 hours | One | a company’s internal tools |
| 99.95% | 4 hours 22 minutes | 30 - 120 minutes | One or more | Online trade, point of sales |
| 99.99% | 52 minutes | 5 - 20 minutes | Two or more | Video delivery |
| 99.999% | 5 minutes | 1 - 60 seconds | Two or more | Financial Transactions |
Use the table above as a reference to create the version tailored to the needs of your applications. Regardless of the recovery strategy chosen, it cannot exceed the Recovery Timeout (RTO).
References