What do uptime and SLA mean?

Uptime and SLA: definition

Uptime states how much of the observed period a service was operational. 99.9 percent uptime per year means, mathematically, about 8.8 hours of downtime; at 99 percent it is about 3.6 days, at 99.99 percent just over 52 minutes. The opposite term is downtime, the time of unavailability.

The service level agreement turns the target value into a contractual obligation. It describes what the provider commits to, how availability is measured, which outages count and what happens if the commitment is missed—often in the form of credits against the fee. Besides availability, SLAs often contain response and recovery times for incidents.

What an SLA should regulate

The definition is decisive: Does availability apply per month or per year? Do planned maintenance windows count as downtime? Is it measured at the provider’s interface or from your customers’ point of view? Depending on the answers, the same percentage can mean very different things. Support hours, around the clock or only on weekdays, belong here too.

The legal consequences also need to be reviewed. Usual credits are typically capped at a portion of the monthly fee and do not compensate for resulting damage, such as lost revenue. An SLA therefore offers above all commitment and predictability, but no substitute for your own measures to ensure reliability.

Real-world example: a car dealership with appointment booking

A car dealership books workshop appointments and test drives through an online application. Two hosting plans are available: 99.5 percent and 99.9 percent availability, with a significant price difference. Mathematically, that is up to about 3.6 hours of downtime per month versus about 43 minutes.

The dealership works out what an hour of downtime actually costs and finds that customers can also call, but during opening hours every outage is annoying. It chooses the higher plan, agrees on announced maintenance windows at night and monitors availability itself so it can document the commitment.

Common mistakes and what to watch for

A common misconception is to equate the host’s commitment with the availability of your own application. Errors in the code, in the database or in third-party systems are not covered by the data center’s SLA. If you chain several services, you must multiply their availabilities: two services at 99.9 percent each yield only about 99.8 percent together.

Nano Concepts helps define suitable availability targets, compare providers and contracts and design the architecture so that the targets are achievable—including monitoring, incident management and regular reports with which you can document the availability of your systems to customers and clients.

• Read the measurement period, method and exceptions in the SLA carefully
• Put maintenance windows and their notice period into the contract
• Check response and recovery times separately from availability
• Know the amount and limits of credits for missed targets
• Set up your own availability measurement as evidence