A service level agreement (SLA) is a formal commitment from a service provider to its customers about how reliably a service will perform, how that is measured, and what the customer receives, such as service credits, if the commitment is missed.
What is an SLA?
A Service Level Agreement (SLA) is a formal agreement between a service provider and its customers that defines expected service levels, how those levels are measured, and the consequences of failing to meet the agreed commitments.
For example:
A cloud service provider commits to 99.5% monthly availability. If availability falls below the agreed level, eligible customers may receive service credits according to the terms of the SLA.
An SLA is the third of three related terms. The SLI measures performance, the SLO is the internal target, and the SLA is the external promise. SLO vs SLA vs SLI compares all three.
What does an SLA include?
An SLA can define:
- The commitment: an availability or performance level, such as 99.5% monthly uptime or responses in under 1 second.
- How it is measured: the measurement period, the data source and the calculation method.
- Exclusions: what does not count, such as scheduled maintenance, customer-caused errors or force majeure.
- Remedies: service credits or other compensation if the commitment is missed, and how customers claim them.
- Responsibilities: what the provider and the customer each have to do, such as reporting an incident within a set time.
SLA example: service credits for missed availability
Many SLAs pay service credits in tiers: the further availability falls below the commitment, the larger the credit. An illustrative tier table for a 99.5% monthly SLA:
| Monthly availability | Service credit |
|---|---|
| 99.5% or above | None: commitment met |
| 99.0% to below 99.5% | 10% of the monthly fee |
| 95.0% to below 99.0% | 25% of the monthly fee |
| Below 95.0% | 100% of the monthly fee |
If a customer pays $20,000 a month and availability for the month is 99.3%, they would be owed a 10% credit: $2,000. Credits are often capped at the monthly fee and usually have to be claimed within a set period.
A 99.5% monthly commitment allows about 3 hours 36 minutes of downtime in a 30-day month. Allowed downtime by SLO lists the figures for other targets.
How is an SLA different from an SLO?
An SLO is primarily used to guide internal reliability engineering decisions. An SLA establishes an external commitment to customers.
| SLO | SLA | |
|---|---|---|
| Audience | Engineering, SRE and product | Customers, sales and legal |
| Typical target | Stricter, such as 99.9% | Looser, such as 99.5% |
| If missed | Reliability work takes priority | Service credits or other remedies |
| Changed by | An internal decision | Contract negotiation |
An SLO breach does not automatically constitute an SLA breach. The SLA may have different thresholds, measurement methods, exclusions, and evaluation periods. SLO vs SLA vs SLI covers the difference in detail.
How do you set an SLA target?
- Start from the SLO, not the other way round. The SLA should be a level the service already meets comfortably, with an internal SLO above it.
- Leave a buffer. A 99.9% SLO behind a 99.5% SLA gives the team room to detect and fix problems before customers are owed credits.
- Check what it depends on. A service cannot commit to more availability than the cloud regions, databases and third-party services it relies on can deliver together.
- Define measurement precisely. State the period, the data source and the exclusions, so the provider and customer calculate the same number.
- Price the remedy. Know what a missed month would cost in credits before signing.
What happens when an SLA is breached?
When an SLA is breached, the remedies in the agreement apply: usually service credits, and in repeated or severe cases, the right for the customer to terminate. Beyond the contract, a breach can affect renewals, reputation and future negotiations.
Because the SLA is usually looser than the SLO, most SLA breaches are preceded by an SLO breach and, before that, by a period of high error-budget burn rate. That earlier warning is the time to act. What an SLO breach costs the business follows the full chain.
Frequently asked questions about SLAs
What is an SLA in simple terms?
An SLA is a promise to customers about how reliably a service will work, such as 99.5% availability each month, with a defined compensation, usually service credits, if the promise is broken.
What is an example of an SLA?
A cloud provider commits to 99.5% monthly availability and pays a 10% service credit if availability falls between 99.0% and 99.5%, rising to larger credits for lower availability.
What is the difference between an SLA and an SLO?
An SLO is an internal reliability target used by engineering. An SLA is an external commitment to customers with contractual consequences. The SLO is usually stricter, so problems are caught before the SLA is at risk.
Is a higher SLA always better?
Not for the provider. A higher SLA means less allowed downtime and larger credit exposure. It should only be offered when the service and its dependencies can reliably exceed it, with an internal SLO above it.
Do all services need an SLA?
No. SLAs are for services sold to or relied on by customers under contract. Internal services usually have SLOs but no SLA.
Key takeaways
- An SLA is a commitment to customers with defined remedies, usually service credits.
- It sets the commitment, how it is measured, exclusions and remedies.
- An SLA should be looser than the internal SLO, leaving a buffer to act.
- An SLO breach does not automatically mean an SLA breach.