SLAs, SLOs, and SLIs: Measuring Service Reliability

In Site Reliability Engineering (SRE), reliability is quantified through three distinct, hierarchical concepts:

1. Service Level Indicator (SLI)

The quantifiable, real-time metric that measures how well a service is performing from the perspective of the user. An SLI is always expressed as the ratio of successful events to total valid events:

SLI = \frac{Good Events}{Total Valid Events} × 100\%

• Example: The proportion of HTTP requests to /api/v1/checkout that return status < 500 with response time < 250ms over a 5-minute window.

2. Service Level Objective (SLO)

The target reliability goal agreed upon by engineering and product management over a defined rolling window (e.g., 30 days).

• Example: "The checkout service SLI must be ≥ 99.9\% over any rolling 30-day window."
• Operational Meaning: The SLO dictates the Error Budget (100\% - SLO) used to balance deployment velocity with system stability.

3. Service Level Agreement (SLA)

The legally binding commercial contract between a service provider and its customers defining promised service availability.

• Example: "We guarantee 99.0% monthly uptime. If availability drops below 99.0%, customers receive a 25% billing credit."
• Financial Consequence: Violating an SLA directly triggers financial refunds, service credits, or contract termination clauses.