Uptime

What it means

Uptime is the share of time a service is available. It is quoted as a percentage for a period and is the most common parameter in an SLA for hosting, cloud services and applications.

The percentages mislead: 99 % a year is more than three days of downtime, 99.9 % is almost nine hours, and 99.99 % under an hour. The price difference between those tiers is usually a multiple.

Uptime measures whether a service responds, not whether the thing a customer needs actually works. A shop’s server can run all day, but if the payment gateway or the login fails, nobody pays. What matters is where availability is measured and whether the key step is checked, not just the home page.

How it works in practice

When assessing a provider, look at what counts towards uptime: planned maintenance windows are usually excluded, though for an online store on a Saturday evening an outage is an outage. Check who measures it and how unavailability is evidenced.

Availability is completed by recovery time after a failure - which is the subject of a recovery plan and a tested backup. High uptime without the ability to restore data quickly is half a solution.

Even a small firm benefits from independent monitoring that tests the site from outside at regular intervals and logs outages. Its record can prove an SLA breach to the provider, and it also gives early warning that a domain or an SSL certificate is about to expire.

What to watch out for

Compensation for breaching an SLA is usually a discount off the monthly fee - a fraction of the real loss from a shop being down. So for critical services the question to settle is the fallback, not the size of the credit.

A service rests on a chain of providers: hosting, domain, payment gateway, the API link to the warehouse. Even if each link is promised high availability, the whole is always less available than its weakest part, and the outages add up rather than average out.