Uptime guarantee
99.9% uptime / SLA uptime guarantee / service level agreement credit
An uptime guarantee is a provider's promise that a service will be available for a stated share of each month, such as 99.9%. In the published terms of large cloud providers, the remedy for a miss is a credit on future bills that the customer has to claim, and that credit is the only remedy.
A percentage close to 100 is easier to judge as time. A 30-day month has 43,200 minutes. At 99.9%, a provider can be down about 43 minutes that month and still meet its promise. At 99.99% the allowance is about four minutes, and at 99.5% it is more than three and a half hours.
Amazon's published terms for its EC2 servers show how the remedy works. A miss earns a service credit, worked out as a percentage of the monthly bill. It applies only against future payments, and AWS says credits do not entitle the customer to any refund. The claim must arrive by the end of the second billing cycle after the incident, with logs that corroborate the outage.
AWS and Google Cloud both call their service level agreement the customer's sole and exclusive remedy. Google requires notice to its support team within 60 days, with log files, or the credit is forfeited. Both list exclusions, such as problems caused by factors outside the provider's reasonable control.
In practice
A company runs its booking site on a single AWS EC2 server that costs $300 a month. The site is down for 12 hours in a 30-day month, so uptime is about 98.3%. That is below 99.0% and above 95.0%, which under the single-server terms earns a 30% credit, or $90 off a future bill. It arrives only if the claim is filed in time with logs, and the bookings lost during those 12 hours are outside the remedy. The figures are a worked example.
Why it matters to you
The uptime number on a hosting quote describes what the provider owes when it fails, and that is usually a small credit. What protects a site that brings in leads is how it is built and watched. AWS offers its higher 99.99% target only when servers run across two or more separate locations, called availability zones.
What to ask or check
- 01What uptime percentage is promised, and how many minutes of downtime a month does that allow?
- 02What is the remedy when the promise is missed, and does a credit have to be claimed?
- 03Who watches the site for outages, and how quickly do they respond?
- 04Does the site run on one server, or on servers in more than one location?
What people get wrong
That an uptime guarantee pays for business lost during an outage. The published AWS and Google Cloud terms make a credit against future bills the sole and exclusive remedy.
Red flags
- An uptime percentage on a quote with no stated remedy, claim process or exclusions.