High availability (HA)

High availability (HA) is a set of design practices aimed at keeping a service running continuously despite the failure of individual infrastructure components. Rather than relying on a single server, disk or network link, an HA architecture duplicates critical parts and automates traffic redirection so users never notice an interruption. Availability is expressed as a percentage — the so-called 'nines' — where 99.9% means a few hours of downtime per year and 99.999% just a few minutes.

How high availability works

HA rests on three pillars. The first is redundancy: eliminating single points of failure (SPOFs) by duplicating servers, power supplies, disk controllers and network paths. The second is failover — a mechanism that detects a fault and automatically reroutes traffic to a healthy node, typically using a load balancer or a floating virtual IP. The third is state synchronization: data must stay consistent across nodes, achieved through database replication and shared or distributed storage.

In practice, servers are joined into a cluster whose nodes continuously exchange heartbeat signals. A missing heartbeat is treated as a failure and triggers a switchover. A key concern is avoiding the 'split-brain' condition, where two nodes both believe they are active and corrupt shared data.

Practical application

High availability is standard wherever downtime causes real losses: online stores, banking systems, SaaS platforms, and DNS or mail servers. Typical deployments include a primary-replica database pair with automatic promotion of the replica, an application cluster behind a load balancer, and geo-redundancy that spreads the service across multiple data centers.

When designing HA, it pays to pair it with defined RTO and RPO targets and with uptime monitoring that measures real availability and alerts on incidents. For VPS and dedicated servers, HA is most often delivered as a cluster of nodes with load balancing and cross-location data replication.