Implementing Service Level Objectives

What the book is about

Most writing on service level objectives explains the idea in a page: pick an indicator, set a target, spend the error budget. This book, published by O'Reilly in 2020, is about what happens after that page. Alex Hidalgo spent years running SLO programmes as a site reliability engineer, and he writes for teams that do not yet have the infrastructure, tooling or culture that the famous examples assume. The book is organised in three movements: developing SLOs (what makes a good indicator, how to choose a target, how to think about error budgets), implementing them (measuring, alerting, dependencies, data and statistics), and building an SLO culture across an organisation. Several chapters are written by guest authors from the reliability community.

Is it worth reading if I have read the Google SRE books?

Yes, and it is the natural next step. The Google books explain why SLOs matter and show how one very large company uses them. This one spends its pages on the questions you hit when you try it yourself: which requests count, what to do with a dependency you do not own, how to set a target when you have no history, how to alert on burn rate instead of on every blip, and how to get product managers to treat the error budget as a real input to planning. If you only read one book about SLOs specifically, make it this one.

Prerequisites and time

You should be comfortable with how a web service handles requests, what latency percentiles mean and what a metrics system such as Prometheus does. No particular cloud or language is assumed. The book is around 400 pages; reading it properly, with notes, takes three to four weeks of evenings. The statistics chapter is the densest part, and it is worth reading slowly rather than skipping.

Free or paid

It is a paid book, sold in print and ebook through O'Reilly and the usual retailers, and included in an O'Reilly online learning subscription. Many universities and employers provide that subscription, so check before you buy. Free companions are the chapters on SLOs in Google's online SRE books.

What you can do after it

• Write SLIs for a real service that reflect what users experience, not what is easy to measure.
• Set an initial target from data or, without data, from a reasoned guess you plan to revise.
• Build multi-window burn rate alerts instead of threshold alerts that page for nothing.
• Write an error budget policy that says what the team actually does when the budget runs out.
• Explain SLOs to non-engineers in terms of user happiness and trade-offs.

Common questions

Is it dated because it is from 2020? Very little of it is tied to specific tools, so it has aged well. The ideas about measurement, targets and culture are the same today.

Do I need to be an SRE to benefit? No. Backend engineers who own a service, platform engineers who build observability, and engineering managers all get value from it.

Best for: engineers who own production services and want to replace vague uptime promises with numbers the whole team can reason about.