Term
High availability
Designing a system to run almost without outages even if one component fails, for example through redundant servers, networks and data. It is measured as a percentage of uptime and becomes part of the SLA.
Updated
What it means
High availability means designing an information system so that it runs almost without outages even when one of its parts fails. It is achieved mainly through redundancy, duplicating critical components such as servers, network connections, power supplies and data storage.
Availability is measured as the percentage of time the system is up. The difference between 99 % and 99.9 % may seem small, but the former means several days of downtime a year and the latter only a few hours. The required level is usually written into the SLA with the service provider.
High availability differs from disaster recovery. High availability tries to prevent an outage by switching automatically to a standby component, whereas disaster recovery deals with restoring service after a serious event, such as a fire in a data centre, when the system stops working for a while.
How it works in practice
Many companies today achieve high availability through cloud services that allow applications and data to be spread across several data centres or zones. With their own infrastructure it means investing in duplicated hardware and a reliable data centre.
The availability level should match how critical the system is. For an online store, a payment system or accounting software used by many clients, downtime is costly, while an internal tool may need a lower level. Higher availability costs more money and effort.
High availability does not replace data backup. If data is corrupted or encrypted by ransomware, the error is quickly replicated to the duplicated systems too. Separate backups and defined RPO and RTO targets are therefore still needed.
What to watch out for
Redundant components only work if failover is tested regularly. Without testing, the standby system may fail to start or lack current data when a real outage occurs. Regular drills should be part of operations.
When assessing a provider's SLA, check how availability is measured, what does not count as downtime, such as planned maintenance, and what compensation applies if the level is missed. Compensation in the form of a fee discount usually does not cover the real damage caused by an outage.
Need the term explained on your own situation? Consulting from €55 / hour excl. VAT, depending on topic and complexity.