8.3. SLOs, SLAs, and Error Budgets
Reliability is not a goal of “no failures” — it is a number you choose, because the last nine costs exponentially more than the one before it. SLOs make that choice explicit: define what “working” means, set a target below 100%, and the resulting error budget becomes the objective currency for deciding when to ship features versus when to stop and fix. This chapter covers SLIs, SLOs, and SLAs, the error-budget model that balances velocity against reliability, and the operational-toil discipline from Google’s SRE practice.
Topics Covered
Section titled “Topics Covered”- 8.3.1. SLI (Service Level Indicator): What Should Be Measured?: Covers choosing service level indicators: the specific measurements that actually reflect user experience.
- 8.3.2. SLO (Service Level Objective): The Art of Setting Targets: Explains setting service level objectives: choosing a reliability target below 100% and defending it.
- 8.3.3. SLA (Service Level Agreement): Legal Obligations: Covers service level agreements, the contractual and legal obligations that back an SLO with consequences.
- 8.3.4. Error Budget: Balancing Innovation and Reliability: Explains the error budget that quantifies acceptable failure and balances feature velocity against reliability.
- 8.3.5. Toil: Measuring and Reducing Operational Burden: Covers toil: measuring and systematically reducing the manual, repetitive operational burden.
- 8.3.6. Practical Lessons from the Google SRE Book: Distills practical lessons from Google’s SRE practice on running reliable systems at scale.