SLA and 99.9% Availability: What It Really Means
By The Atlas Hoster team4 min read
SLA: Definition
A Service Level Agreement (SLA) is a contract that defines the level of service a hoster commits to delivering, along with the compensation provided if it fails to do so. Without a written SLA, the "99.9% uptime" promise displayed on a homepage has no value: it defines neither what is measured, what is excluded, nor what you are owed in the event of a breach.
A serious SLA contains at least three things: the metric being measured, the reference period, and the credits in case of non-compliance. Read these three points before comparing offers. A hoster who refuses to write them down is indirectly telling you that it does not wish to be held accountable.
99.9%: How many outages per month
Uptime is expressed as the percentage of operational time over a given period. In a 30-day month, that is 43,200 minutes:
| Uptime | Allowed downtime per month |
|---|---|
| 99.9% | approx. 43 minutes |
| 99.0% | approx. 7 h 12 |
| 95.0% | approx. 36 hours |
Therefore, a 99.9% SLA does not mean "zero outages": it allows for approximately 43 minutes of downtime per month without triggering a credit. This is the order of magnitude to keep in mind when comparing "guaranteed uptime" offers. The difference between 99.9% and 99.99% is a factor of ten: 43 minutes versus just over 4 minutes. This is why the digit after the decimal point matters, and why some hosters highlight it in their premium offers.
What is measured and what is excluded
Look closely at what the agreement covers. An honest SLA specifies that it measures the infrastructure: the network, hypervisor, storage, and power supply. It also lists what is excluded. For example, Atlas Hoster’s agreements exclude maintenance announced at least 48 hours in advance, incidents caused by customer content or configuration, attacks exceeding subscribed protections, force majeure, and suspensions for non-payment or illicit use.
These exclusions are not a trap: they are the industry standard. What matters is that they are written and reasonable, not discovered retroactively. An SLA without exclusions, or with exclusions so broad that they void the agreement, should alert you. The reference period also matters: a "monthly" commitment applies to a rolling month, not an entire year. A hoster displaying 99.9% without specifying the period is not committing to anything measurable.
GTI and GTR: Response and Recovery Time
The SLA also covers response times. The GTI (response time) runs from the alert to the start of action; the GTR (recovery time) covers the return to an operational state, potentially degraded. At Atlas Hoster, the tiers range from 4 business hours (Essential tier) to 15 minutes 24/7 with recovery within 2 hours (Critical tier).
Keep in mind that the GTR does not promise the resolution of the root cause: it promises that the service will function again. The permanent fix may take longer and be communicated separately. This is an important distinction: do not confuse "the site is back" with "the problem is fixed."
Credits in case of non-compliance
A useful SLA specifies what happens when the commitment is not met. Atlas Hoster applies tiered credits: 10% of the monthly fee if uptime drops below 99.9%, 25% below 99.0%, 50% below 95%, and 10% per response time breach, up to a limit of 50% of the monthly fee. The credit is applied to the next invoice.
Check two points: the deadline to request the credit (often thirty days) and the fact that the credit constitutes the sole compensation. These limits are standard, but they must be known. A credit capped at half the monthly fee means the hoster does not reimburse your lost business revenue: that is the rule of the game, and it must be understood before relying on it. Beyond credits, an SLA also serves as a signal: a hoster that commits in writing with precise figures has an interest in honoring its commitments. This is an indicator of seriousness, just like support quality.
Verify availability yourself
Do not rely solely on the SLA: verify it. A good hoster provides a public status page and, within the client portal, the availability history of your services. You can also monitor your site yourself using a testing tool that pings your page at regular intervals and alerts you in case of failure.
The simple reflex: ask the hoster where to check the actual availability of the past month. A hoster that measures and publishes its own figures allows you to verify its claims. To track your incidents and requests, Atlas Hoster offers to read your monthly reports and follow your support requests.
A 99.9% SLA is a contractual floor, not a guarantee of perfection. Understanding its figures and exclusions allows you to make an informed choice. To learn more, read our guide on unmanaged VPS, our guide on high availability, and our guide on 24/7 managed hosting.
Consult the Atlas Hoster service commitments to view a complete SLA, complete with figures and exclusions.
FAQ
99.9% vs. 99.99%, what is the difference? A factor of ten: approximately 43 minutes of allowed downtime per month versus just over 4 minutes. The digit after the decimal point has a real cost.
Does maintenance count as an outage? Generally no, if it is announced in advance (48 hours at Atlas Hoster). This is a standard exclusion from the uptime calculation.
What to do if the hoster does not respect its SLA? Check the credit conditions (request deadline, cap) and request the credit within thirty days, as stipulated by Atlas Hoster’s commitments.