Real Organizations Use More Than One Model
In practice, mature engineering organizations don't pick a single model and apply it everywhere. They layer models based on service criticality, team maturity, and available SRE headcount.
A common hybrid at a mid-to-large company:
- Platform SRE team (always on): Owns the shared infrastructure. Serves all 200+ developers. Defines the Golden Path.
- Embedded SREs (for critical services): 2-3 SREs permanently embedded in the Payments and Identity teams, because those services are existential to the business.
- Consulting SRE pool (for interventions): 4-5 senior SREs available to be deployed to teams in crisis or teams launching new services.
How to decide which model to apply where:
Use service criticality and team maturity as your two axes. A high-criticality, low-maturity team needs a Consulting or Embedded SRE immediately. A low-criticality, high-maturity team needs only the Platform. A high-criticality, high-maturity team may need only light-touch Embedded support to maintain standards.
The key principle: SRE resources are finite and should be allocated where reliability risk is highest. Don't embed an SRE in a team that is already meeting its SLOs. Don't leave a team in crisis to figure it out alone because they weren't on this quarter's rotation.