Availability is usually expressed as a percentage of time a service is up. The differences between the numbers look small, but they are not.
The arithmetic
Allowed downtime per year:
- 99%: about 3 days 15 hours.
- 99.9%: about 8 hours 46 minutes.
- 99.95%: about 4 hours 23 minutes.
- 99.99%: about 52 minutes.
- 99.999%: about 5 minutes.
Per month, 99.9% allows roughly 43 minutes; 99.99% allows about 4 minutes.
Every nine costs more
Going from 99.9% to 99.99% is not a small improvement. Four minutes a month is less time than it takes most people to notice a problem, investigate and act. It requires redundancy at every layer, automatic failover, and no single server, link or person whose absence causes an outage. That costs real money.
How to read an SLA
- What is measured? The hardware, the network, the hosting platform, or your application?
- How is it measured, and from where?
- What is excluded? Planned maintenance, problems outside the provider's control, issues caused by your own configuration.
- What is the remedy? Usually service credits, a percentage of the monthly fee. Credits rarely cover the business cost of an outage, so an SLA is not insurance.
Your availability is a chain
Your service is only as available as its weakest dependency: DNS, the network, the server, the database, third-party APIs, and the people who respond. Multiplying several 99.9% components together gives a lower overall figure. Design for the target you need at every link.
Choosing your target
Ask what an hour of downtime costs your business, at which times, and for which services. A public website may be fine at 99.9%. A payment platform or a system your customers depend on may justify the cost of 99.99%.
How IDX helps
Our team can help you work out the availability your services really need and design for it, from redundant servers and tested failover with IDX Disaster Recovery to monitoring by our NOC. Planned maintenance on IDX services is announced on our status page in advance.
IDX on digital infrastructure
Back to blog →