High Availability Server Architectures: Keeping Services Online

High Availability Servers

High availability (HA) is designing a service so that it keeps running when a component fails. Instead of relying on one server, HA spreads the work and the risk across several, so a failed drive, power supply, server or even a whole rack does not take the service down. Refurbished servers make this kind of redundancy far more affordable.

Quick answer: high availability removes single points of failure at every layer: redundant parts inside each server, at least two servers per critical service (active-active or active-passive), shared or replicated storage, redundant network and power paths, and automatic failover with health checks. Test failover regularly so you know it works.

Common high availability architectures

ArchitectureHow it worksGood for
Active-passive failoverA standby server takes over when the primary failsDatabases, file servers, applications that cannot run on two nodes at once
Active-active with load balancingSeveral servers share live trafficWeb applications and stateless services
Hypervisor clusterVMs restart on another host if one failsSmall businesses running several workloads
Database replicationPrimary and replica databases kept in syncCritical databases
Multi-siteServices run or can fail over to a second locationProtection against site-wide outages
load balancing in high availability servers

Remove single points of failure, layer by layer

LayerRedundancy to add
Server componentsRAID, dual hot-swap power supplies, redundant fans, ECC memory
ServersTwo or more nodes per critical service
StorageShared storage with redundant controllers, or replicated storage
NetworkTwo switches, teamed server ports, redundant uplinks
PowerTwo power feeds or circuits, UPS, generator or inverter
InternetTwo providers or a backup link
Pro tip: connect each server’s two power supplies to different circuits or UPS units. Dual supplies on the same plug protect against a PSU failure but not a tripped breaker.
Watch out: a cluster that has never been tested is not highly available. Pull a power cord or shut down a node during a maintenance window and confirm services fail over as expected.

Building HA with refurbished servers

  1. Choose identical refurbished servers so any node can replace another.
  2. Use a hypervisor cluster (for example Proxmox VE, Hyper-V failover clustering or VMware) or application-level clustering.
  3. Provide shared or replicated storage that every node can reach.
  4. Add load balancing for web services. See load balancing with refurbished servers.
  5. Monitor every node and set alerts. Our monitoring guide compares tools.
Good to know: two refurbished servers in a cluster often cost less than one new server, which makes real redundancy realistic for small businesses that would otherwise run on a single machine.

High availability reduces downtime from everyday failures; disaster recovery covers the big events. Plan both, using our disaster recovery planning guide and our explanation of uptime SLAs.

Frequently asked questions

What is a high availability server?

A server set up as part of a redundant design, usually a cluster, so that the service continues if the server or one of its components fails.

How many servers do I need for high availability?

At least two for the service itself. Some cluster technologies need three nodes, or two nodes plus a witness, to avoid split-brain situations.

What is the difference between high availability and disaster recovery?

High availability keeps services running through component failures; disaster recovery restores services after major events such as fire or ransomware.

Can refurbished servers be used for high availability?

Yes. Identical refurbished enterprise servers make cost-effective cluster nodes, as they include redundant power, RAID and remote management.

Building a cluster? Configure matching refurbished servers or ask us for volume pricing.

Server Site