High availability

High availability is the ability of a system, network or service to stay operational and reachable, without significant interruption, even when something fails.

In security terms it belongs to the third leg of the classic triad. Confidentiality and integrity get most of the attention, and availability is the one an attacker goes after when the goal is to stop you working rather than to read your data.

Getting there means designing for failure instead of hoping to avoid it, which is a different engineering posture from simply buying more reliable hardware.

How it is achieved

Redundancy. Critical components are duplicated, servers, network paths and storage, so that if one fails another takes over without the service stopping. In practice that means clustering, automated failover and workloads spread across nodes.

Continuous monitoring. Problems have to be seen before they become outages. Automated monitoring, alerting and trend analysis are what let an operations team act while there is still time to act.

Disaster recovery. Plans and procedures for restoring systems and data after a catastrophic event, whether that is a cyberattack, a hardware failure or a natural disaster.

Geographic distribution. If everything depends on one building or one region, the availability of the service is the availability of that place.

A worked example

A company runs an e-commerce platform on a cloud provider. Online transactions are the business, so an outage is not an inconvenience, it is lost revenue.

They deploy multiple web server instances and distributed databases across more than one geographic location, and put continuous monitoring in front of them.

If a server fails, or demand spikes, load is redistributed automatically to the instances still healthy, and the service stays up.

What makes that work is not the architecture diagram. It is that the failover path is exercised regularly, because a failover nobody has tested is a hypothesis.

High availability, backup and disaster recovery

Three things that get bought as if they were one, and they answer three different questions.

High availability answers: what happens when a component fails right now. The answer is that another one takes the load and nobody notices. It protects against failure, and it does not protect against corruption, because a corrupted record replicates to the standby as fast as a correct one.

A backup answers: how do we get back a state from before something went wrong. It is the only one of the three that helps against ransomware and against a bad deployment, and only if the copy is isolated from whatever went wrong, which is why immutable backups exist.

Disaster recovery answers: what happens if we lose a whole site, and how long we can be down for. Its two numbers are the recovery time objective and the recovery point objective, and both of them should come from a business impact analysis rather than from a preference.

Replication is not backup. That sentence is the one worth keeping: a cluster that mirrors data in real time also mirrors the deletion.

Where to read more

TechTarget, High availability: an overview of the concept, its components and how it is applied in different environments.

IBM, What is high availability: use cases, practices and the technology commonly used to keep services running.