Availability means that information and the systems supporting it can be accessed when needed. It is one of the three goals of the CIA triad, alongside confidentiality and integrity. A denial of access threatens availability even if no information is disclosed or altered. section 1.2
High availability (HA) is the aim of keeping critical systems reliable despite failures or maintenance. In a data center, redundancy provides duplicate components so another can take over when one fails. The practical question is not whether downtime matters, but how much the organization can tolerate. section 13.2.1
Common ways to protect availability include:
- Providing independent power feeds, UPS batteries, and generators if power fails. section 13.2.1
- Keeping redundant Internet access and extra servers ready to take over. section 13.2.1
- Using failover, backups, load balancing, or auto-scaling to counter denial of service. section 1.2
- Performing rolling updates so part of a system stays live during maintenance. section 13.2.1
Choosing a downtime target
At 99.99% availability, a system may still be unavailable for about 52 minutes a year. That could interrupt users or transactions. Raising the target to 99.999% reduces the allowance to about five minutes, but demands much greater cost and complexity. Choose a target against the harm downtime would cause, rather than assuming more “nines” are always worthwhile. section 13.2.1
|
Availability
|
Nickname
|
Maximum downtime per year
|
|
99%
|
“Two nines”
|
About 3.5
days
|
|
99.9%
|
“Three
nines”
|
About 9
hours
|
|
99.99%
|
“Four nines”
|
About 52 minutes
|
|
99.999%
|
“Five
nines”
|
About 5
minutes
|
Compare availability percentages with their approximate maximum downtime per year. section 13.2.1
























