The SRE Mindset (Site Reliability Engineering)

This document explores the fundamental concepts of Site Reliability Engineering (SRE), focusing on the triad that sustains the reliability of modern systems: SLIs, SLOs, and Error Budgets.

1. What is SRE?

Site Reliability Engineering (SRE) is a discipline that incorporates aspects of software engineering and applies them to IT operations problems. The goal is to create ultra-scalable and highly reliable software systems.

As the saying goes at Google: “SRE is what happens when you ask a software engineer to design an operations function.”

2. Service Level Indicators (SLIs)

Service Level Indicators are quantitative measures of the level of service being provided.

Definition: A direct measurement of a service’s behavior.
Common Examples:
- Latency: Time taken to service a request.
- Availability: Fraction of the time that the service is usable.
- Throughput: Number of requests processed per second.
- Error Rate: The percentage of requests that failed.

Pro Tip: A good SLI should clearly indicate whether the user is happy or not.

3. Service Level Objectives (SLOs)

Service Level Objectives are target values or ranges of values for a service level that is measured by an SLI.

Definition: SLI <= Target (or SLI >= Target).
Importance: SLOs define what is “acceptable” for the business and the users.
Example:
- “99.9% of database read requests must complete in less than 200ms over a 30-day window.”

4. Error Budgets

The Error Budget is the direct consequence of the SLO. It represents the amount of failure that can be tolerated before the SLO is violated.

The Formula

Error Budget = 100% - SLO

If the Availability SLO is 99.9%, the Error Budget is 0.1%.
What is it used for?
- Innovation vs. Stability: If you have plenty of “budget,” you can launch new features quickly. If the budget is nearly exhausted, you must focus on stability and fixes.
- Alignment: It unites Development (Dev) and Operations (Ops) teams under a common objective.

5. Summary of the Relationship

Concept	Question Answered	Example
SLI	”What are we measuring?”	HTTP 200 Success Rate.
SLO	”What is the ideal value?“	99.5% success per month.
Error Budget	”How much can we fail?“	0.5% allowed failures.

6. Best Practices

Less is More: Don’t try to measure everything. Choose the 3-5 most critical indicators for the user experience.
Don’t Aim for 100%: Perfection is impossible and extremely expensive. Users usually don’t notice the difference between 99.9% and 100% (due to their own internet reliability).
Blame-Free Culture: When an SLO is violated, the focus should be on system improvement, not on finding someone to blame.