Disaster Recovery Metrics
Unit: 7
Lesson: 4
i need a number that tells me how good we are at not imploding
We love a number that directly states how reliable a service on the Internet is.
Availability
The percentage of time that the system is online over a set amount of time, usually a year. The opposite of availability would be downtime. The main metric for downtime is called Maximum Tolerable Downtime (MTD), and it defines the upper limit for downtime for a service. Downtime is measured as the sum of all time it takes for scheduled maintenance and unplanned outages.
| Availability (%) | Annual MTD |
|---|---|
| 99 | 87h 36m |
| 99.9 | 8h 45m 36s |
| 99.99 | 52m 34s |
| 99.999 | 5m 15s |
| 99.9999 | 32s |
| Systems that have almost no scheduled downtime or outages are extremely rare. They are said to have continuous availability. Continuous availability is needed not only in industrial settings, but others where systems failure could bring about injury or loss of life. Such settings include: |
- life-support systems
- air traffic control systems
- communication satellites
- networked self-driving vehicles
- smart traffic signaling systems
Remember how MTD is the final upper limit for how long a service can remain unavailable due to a scheduled maintenance or unplanned outage? There's two more metrics. There's the Recovery Time Objective (RTO) which states how long a service remains unavailable following a disaster. RTO includes the time it takes to find out what happened, find out if it's an issue, perform backups and/or switch to a backup system.
There's also the Work Recovery Time (WRT), which states how long it takes to reintegrate systems and other extra work to restore the service to working, operational condition.
[!]The RTO and the WRT must not exceed the MTD!
There's also the Recovery Point Objective (RPO), which states how much data loss is acceptable for the system, measured in time. For example, if a virus destroyed a database, an RPO of 24 hours means that we have a backup from a maximum of 24 hours before the loss, but not anything farther back than that. Any data lost between the RPO and the present needs to be accepted as a loss or reconstructed. Pray you were using RAID.
#Netplus