Cloud Architecture

The Life Cycle of an SLA in Cloud Computing

The life cycle of a cloud SLA, from defining service levels to monitoring violations and claiming credits, plus what uptime percentages mean in downtime.

On this page
  1. What is an SLA in cloud computing?
  2. The life cycle of a cloud SLA
  3. SLI, SLO, and SLA: how they fit together
  4. What uptime percentages mean in downtime
  5. Composite SLAs
  6. What does a 4-hour SLA mean?
  7. Common misconceptions
  8. Negotiation tips
  9. Key takeaways

What is an SLA in cloud computing?

A service level agreement (SLA) is a contract between a cloud provider and a customer. It states the level of service the provider commits to, how that service is measured, and what happens if the provider falls short. A typical cloud SLA covers:

  • Service description: which services are covered and their scope.
  • Service level objectives: targets such as availability or response time.
  • Metrics and measurement: how availability is calculated and reported.
  • Responsibilities: what the provider and the customer must each do.
  • Exclusions: situations the SLA does not cover, such as scheduled maintenance, customer misconfiguration, force majeure, and beta services.
  • Reporting and remedies: how performance is reported, and what compensation applies, usually service credits.

The life cycle of a cloud SLA

Sources describe the life cycle in different ways. A common six-stage version is:

  1. Discover and compare providers.
  2. Define your requirements.
  3. Establish the agreement.
  4. Monitor for violations.
  5. Claim remedies.
  6. Terminate or renew.

Another common version has five phases: develop a template, negotiate, implement, execute and monitor, and assess and terminate. Both versions cover the same ground.

1. Discover and compare providers

Compare providers on the services you need, the published SLA terms for each, and the procedures for measuring performance and claiming credits. The headline uptime figure is only part of the picture.

2. Define your requirements

Before you negotiate or accept terms, decide what you need. Consider when the service is used, how much downtime the business can tolerate, how much notice you need for maintenance, and how you want to be told about incidents. An uptime target should match the hours you actually depend on the service.

3. Establish the agreement

Standard public-cloud SLAs are usually not negotiated, so read the published terms carefully before you rely on them. Larger contracts may allow negotiation of escalation paths, remedies for missed targets, the formula used to calculate uptime, and exit rights.

4. Monitor for violations

Measure the service yourself where you can. Provider reports may not match what your users experience. Keep dated records of incidents so that you can check them against the SLA’s definitions.

5. Claim remedies

Service credits are the usual remedy. They are often tiered, so larger shortfalls earn larger credits, and they are usually capped. Credits are generally not automatic. Customers normally have to file a claim within a set window after the incident. Credits are not cash refunds, and they rarely cover the business cost of an outage.

6. Terminate or renew

At the end of a term, review the service’s performance history and decide whether to renew, renegotiate, or leave. Check the exit terms, such as notice periods and data export, before you need them.

SLI, SLO, and SLA: how they fit together

  • SLI (service level indicator) is the measured value, such as the percentage of requests that succeed.
  • SLO (service level objective) is the target for that measurement, such as 99.9 percent of requests succeeding.
  • SLA (service level agreement) is the contract. It attaches consequences, such as service credits, to a missed target.

What uptime percentages mean in downtime

Availability percentages can look similar, but they allow very different amounts of downtime. The figures below use an average month of about 30.4 days.

Availability Downtime per month Downtime per year
99.9% about 43.8 minutes about 8.76 hours
99.95% about 21.9 minutes about 4.38 hours
99.99% about 4.4 minutes about 52.6 minutes
99.999% about 26 seconds about 5.26 minutes

Each additional “nine” reduces the allowed downtime tenfold. Whether a target is worth paying for depends on what an outage costs you.

Composite SLAs

Many applications depend on several services at once. If each service must be available for the application to work, the overall availability is roughly the product of the individual figures. Two services that each promise 99.9 percent, used together, give about 99.8 percent, or roughly 88 minutes of downtime a month. This estimate assumes that the failures are independent, which is not always true.

Composite availability also explains why disaster recovery and redundancy matter. Adding redundant components can raise the availability of the whole system beyond what any single service promises.

What does a 4-hour SLA mean?

A “4-hour SLA” usually refers to a commitment to respond to, or resolve, a support issue within four hours. The exact meaning depends on the contract, so check how the provider defines the response time, the resolution time, and the priority level it applies to. Some providers label support issues P1 to P4 by severity, and each provider sets its own targets for each level.

Common misconceptions

  • An SLA is not a guarantee that a service will stay up. It defines the provider’s commitment and the remedy if that commitment is missed.
  • Service credits are not cash refunds, and they rarely cover the business cost of an outage.
  • Durability and availability are different. Availability describes whether a service is reachable. Durability describes whether stored data survives. Amazon S3 Standard, for example, has a 99.9 percent availability SLA and is designed for 99.999999999 percent (eleven nines) durability.
  • Credits are not automatic. Customers usually need to claim them.

Negotiation tips

When you can negotiate, focus on terms that affect you in practice:

  • Align the provider’s maintenance windows with your low-usage periods.
  • Make sure the provider’s support is staffed when your business needs it.
  • Define how uptime and response times are calculated, including what counts as an outage.
  • Ask for regular service reviews and clear escalation steps.
  • Read the list of exceptions closely, and check that exclusions do not cover your normal operations.

Key takeaways

  • An SLA is a contract with a measured target and a stated remedy.
  • Read the exclusions and claim procedures, not only the headline percentage.
  • Use the downtime table to judge whether a target fits your needs.
  • An SLA does not replace a disaster recovery plan, and capacity choices affect what you can commit to. See resource provisioning.

The complete guide to cloud architecture

Cloud Computing Architecture: The Core Components
Read the guide

Keep exploring

Guides from other topics

All topics →