SLIs vs SLOs vs SLAs: How Reliability Is Measured, Targeted, and Promised

GCP Study Hub
October 19, 2026

SLIs, SLOs, and SLAs are three related terms that describe how a company measures the reliability of a service, what level of reliability it aims for, and what it promises its customers. They build on each other in order. The service level indicator (SLI) is the measurement. The service level objective (SLO) is a target defined on top of the SLI. The service level agreement (SLA) is a contract built on top of the SLO. For the Professional Cloud Architect exam, it helps to keep these layers distinct and to understand why each one is set where it is.

SLI: what actually happened

An SLI is a direct measurement of how a service behaves. It is the actual data you look at to see how the service performed, and it is usually tracked as the rate of successful requests compared to failed ones. For example, an SLI might show a monthly uptime of 99.97 percent. That number on its own does not say whether the service did well or badly. It only records what happened.

Uptime is not the only option. Latency, error rate, and throughput are other common SLIs, and the right choice depends on what the users of the service care about. For a user-facing API, latency and error rate are usually the best fit, because users notice slow or failed responses. For a batch data pipeline, throughput tends to matter more.

SLO: what counts as good enough

An SLO is the target value or range for an SLI. This is where a team decides what level of performance is good enough. The SLI tells you what happened, and the SLO tells you whether that result is acceptable. Continuing the example, suppose the company sets an internal target of 99.99 percent uptime. Measured against that target, the 99.97 percent SLI falls short, so the service missed its SLO that month.

When setting an SLO, the goal is the lowest level of reliability that users will accept, not the highest number possible. Higher reliability costs more to run. Every extra nine requires more redundancy and more engineering effort, and it slows down development. A realistic SLO is one that matches what users actually need.

SLA: what is promised, with consequences

An SLA is a contract with customers that attaches penalties when the service misses the reliability level written into the agreement. The penalty is usually a partial refund or extra service time at no cost. The difference from an SLO is that an SLO is an internal goal, while an SLA has consequences attached. For example, the SLA might promise credits if uptime drops below 99.9 percent.

The SLA target is usually looser than the internal SLO, and the gap between them acts as a buffer. When the service drops below its internal SLO, the team knows something is wrong and can fix it before the contract terms are breached. In the running example, 99.97 percent missed the internal target of 99.99 percent but stayed above the 99.9 percent SLA threshold, so no credits are owed.

The three numbers side by side

  • SLI: measured monthly uptime of 99.97 percent, which is simply what happened.
  • SLO: an internal target of 99.99 percent, which was missed, so the team knows to investigate.
  • SLA: a customer commitment of 99.9 percent, which was met, so no credits are owed.

How SLAs shape architecture decisions

Google Cloud uses the same model for its own services. Spanner, for example, offers an SLA of up to 99.999 percent uptime, which works out to about five minutes of downtime per year. If Spanner does not meet that commitment, customers receive financial credits.

This carries over directly into system design. When a system needs a certain level of uptime, the services it runs on have to support that target. That is why the SLAs of Google Cloud services matter when choosing which ones to build on, which is part of the service selection a Professional Cloud Architect works through when designing a system.

Our Professional Cloud Architect course covers SLIs, SLOs, and SLAs alongside Spanner and high availability design, with practice questions that drill these distinctions.

Get tips and updates from GCP Study Hub

arrow