The term high availability is used for computer systems that are optimized for
non-stop operation with a minimum of unplanned downtime. Usually this is
achieved by avoiding single points of failure and adding additional redundancy
to the entire system. Common solutions are hot-standby systems that take over
the role of a crashed server in a split second. More sophisticated solutions
offer both scalability and high availability by using
entire clusters of servers that run their applications in a distributed
manner but look like one virtual
server to their clients.
But how high is HA? Have a look at our table or enter your own percentage to
see how much unplanned downtime per year or month has to be tolerated at a
certain service level.
There are some common misunderstandings about HA.
Take care of this, especially when negotiating contracts.
- Planned downtime for maintenance (e.g. upgrading the hardware,
installing new software) is usually not included in the
guaranteed maximum downtime. That's why it's called unplanned downtime.
- An availability of n% per year is not the same like n% per month, since
it can not be guaranteed that the entire downtime per year will be evenly
distributed over all twelve months. Example: to get an availability of 97%
per month you need approximately 99.75% per year in the worst case.