Google SRE Book — Introduction
The canonical first-party text defining Site Reliability Engineering (site-reliability-engineering), free online from Google. Ingested to give the spoke’s SRE page a definitional source beyond the applied google-sre-agentic-ai case.
Citable claims
- Origin & definition: SRE began at Google in 2003 when Ben Treynor Sloss was asked to run a seven-person production team; his definition is “what happens when you ask a software engineer to design an operations team” — operations treated as a software problem.
- The 50% cap on toil: aggregate operational work (tickets, on-call, manual tasks) is capped at 50% of an SRE’s time; the rest must go to engineering. This deliberately breaks the link between service growth and headcount, forcing automation and self-healing systems.
- Error budget: permitted unavailability derived from the reliability target — a 99.99% service has a 0.01% downtime budget. It resolves the structural tension between dev (ship features fast) and ops (keep it stable): launches spend the budget; exhausting it freezes risky changes.
- Change is the main risk: ~70% of outages stem from changes to a live system → progressive rollouts, fast problem detection, and safe rollback are core practices.
- On-call sanity: target a maximum of ~2 events per 8–12h shift, so each gets a real response + postmortem (pager-fatigue guardrail).
Tier
T1 — first-party (Google’s own SRE Book). Foundational/definitional, not a dated snapshot.
Related
site-reliability-engineering · service-level-objectives · platform-ops · aiops · google-sre-agentic-ai