What is SRE?¶
Site Reliability Engineering applies software engineering to operations. Reliability becomes a feature that is measured, budgeted and engineered, instead of something achieved by heroics.
Core principles¶
- Reliability is a product feature. Users feel availability and latency, not your architecture.
- Measure it with SLOs. Agree on a target for what "reliable enough" means.
- Spend an error budget. The gap between the SLO and 100% is room for change and risk.
- Eliminate toil. Automate repetitive manual work.
- Learn from failure. Run blameless postmortems and fix systemic causes.
- Keep it simple. Complexity is the main enemy of reliability.
SRE, DevOps and platform engineering¶
| Main question | Typical output | |
|---|---|---|
| DevOps | How do we ship quickly and safely together? | Culture, CI/CD, shared ownership |
| SRE | Is the service reliable enough, and how do we prove it? | SLOs, error budgets, on-call, postmortems |
| Platform engineering | How do we make good practice the easy path? | Internal platform and golden paths |
They overlap heavily, and good teams use all three ideas.