What is high availability?
High availability (HA) is the principle of designing systems that tolerate failures without going down. Hardware fails, disks die, power disappears, and updates require reboots. These are not exceptions, they are operating conditions. A highly available system is built so that none of those events reaches the users: operations continue on the redundant component while the failed one is repaired.
The basic recipe has two ingredients: redundancy, meaning that everything critical exists in at least two copies (servers, networking, power, internet connections), and automatic failover, which detects the failure and moves the traffic without waiting for a human.
The nines: what does uptime cost?
Availability is measured as uptime in percent ("the nines"):
- 99% ≈ 3.7 days of downtime per year
- 99.9% ("three nines") ≈ 8.8 hours per year
- 99.99% ≈ 53 minutes per year
- 99.999% ("five nines") ≈ 5 minutes per year
Each additional nine requires more redundancy, more automation and more testing, and the price rises accordingly. So the right question is not "how many nines can we get?" but "what does one hour of downtime cost us per system?". The webshop and the finance system rarely have the same answer. The requirement is set per system and written into your SLA, so the delivery can be measured.
High availability vs. disaster recovery
The two are often confused, but they solve different problems:
- High availability protects against everyday component failures: a dead server, a network fault, a reboot. The reaction is automatic and measured in seconds.
- Disaster recovery handles the catastrophe: ransomware, fire, loss of an entire data centre. The reaction follows a plan and is measured in hours.
The point: HA replaces neither backup nor DR. A fully redundant system also replicates ransomware-encrypted data in real time: perfectly available and perfectly ruined. You need both, dimensioned according to the criticality of the systems.
How high availability is achieved in practice
In modern operations the tools are readily available: server clusters and virtualisation with automatic failover, load balancers in front of web applications, redundant internet connections and power supplies, and in cloud platforms such as Azure built-in availability zones, where workloads are spread across physically separate data centres. At least as important is monitoring: a failure on the redundant component must be detected immediately, otherwise you are running without a safety net without knowing it.
How MI Support IT can help
We design and operate highly available environments for Danish SMBs: redundant IT hosting and cloud solutions, round-the-clock monitoring and clear SLAs, combined with the backup and contingency plan that catch you when disaster strikes anyway. Contact us if downtime has started to cost real money.