High Availability (HA) keeps an IT service accessible through selected failures using redundancy, health monitoring and failover to a working resource.
Simple definition
High Availability (HA) refers to mechanisms designed to keep an IT service accessible despite certain failures. It can rely on multiple servers, network paths or storage components, combined with health monitoring and a way to switch to a healthy resource. HA aims to reduce service interruptions; it does not guarantee perfect uptime.
What is it used for?
Maintain access to business-critical systems during selected hardware or software failures, reduce disruption during maintenance and support uptime targets. The protected scope matters: having two servers is not sufficient if both still depend on one power feed, firewall, storage controller or Internet connection.
How does it work?
HA platforms use health checks or periodic communication (heartbeats) to detect failures. When a component fails, failover can redirect traffic, promote a standby instance or restart workloads elsewhere. Designs may be active-passive or active-active. Where cluster members coordinate access to shared data, quorum or an equivalent safety mechanism helps reduce split-brain risks.
- Monitoring Health checks and heartbeats
- Failure Node unavailable
- Failover Cluster recovery
- Surviving node Spare capacity
- Service restored Possible downtime
HA reduces downtime; depending on the platform, VMs may need to restart and service recovery can take time.
Key points
- A hidden single point of failure in storage, networking, power, DNS or security gateways.
- Insufficient spare capacity to absorb workloads after a node fails.
- Faulty quorum or heartbeat configuration, causing failed or unsafe failover.
- Application or shared-storage inconsistency after an incident.
- Failover never tested, leaving the actual recovery time unknown.
Points to watch
A cluster of virtualization hosts uses a storage design suitable for availability. If one host fails, the VMs affected may be restarted on surviving hosts if sufficient capacity exists. That does not necessarily mean the guest workloads continue running without any interruption.
Active-passive, active-active and fault tolerance
In an active-passive setup, a standby component takes over if the primary fails. Active-active architectures distribute selected workloads across several instances but require careful data consistency, session handling and load balancing. Capacity planning must evaluate how the service behaves with a node missing rather than only when every member is healthy.
Fault-tolerant designs can target even shorter interruptions than some HA mechanisms, usually at additional complexity or cost. The word 'high availability' should not be interpreted as a promise of zero downtime.
Failure detection, quorum and failover
Health checks monitor whether components can serve requests. Failover may mean switching a virtual IP address, adjusting load balancers, promoting a replica or restarting a VM on another host. The actual duration depends on detection time and application dependencies.
Quorum helps some clusters avoid conflicting decisions after a network partition. Without appropriate split-brain protection, two separated nodes may both believe they own the same resource and corrupt shared data.
HA versus disaster recovery, replication and backups
High availability primarily addresses service interruptions caused by failures that the architecture can tolerate. Disaster recovery planning provides procedures for restoring service after major incidents. Replication may maintain an accessible secondary copy, while backup stores recoverable historical versions.
An HA cluster cannot undo an accidentally deleted database or stop an attack affecting all nodes. Service Level Agreements (SLAs), RTO and RPO should therefore be measured and validated through realistic testing.
Related terms
In short
Key takeaway: HA combines redundancy, monitoring and failover to limit outages. It must address real single points of failure and complement disaster recovery planning and backups.