SLI/SLO Management

SLI/SLO Management: Under the Hood

Mathematical Modeling of SLIs

Service Level Indicators (SLIs) must be quantifiable fractions of good events over total events.

• Availability SLI: $SLI_{avail} = \frac{\text{Successful Requests}}{\text{Total Valid Requests}}$
• Latency SLI: $SLI_{latency} = \frac{\text{Requests } < 200ms}{\text{Total Requests}}$
• Continuous Aggregation: SLIs are typically evaluated over rolling windows (e.g., 28 days) using integration of time-series metrics.

flowchart TD
%%{init: {"theme": "default", "flowchart": {"useMaxWidth": true}}}%%
    subgraph ModelingSLIMathematics ["SLI Mathematics<br><br><br>"]
        Events[Raw Telemetry] -->|"Aggregate(1m)"| Metric[Prometheus Metric]
        Metric -->|"Calculate(Good/Total)"| SLI[SLI Value]
    end
    subgraph ErrorBudgetErrorBudgetPolicies ["Error Budget Policies<br><br><br>"]
        SLI -->|"Compare(SLO)"| Budget[Burn Rate]
        Budget -->|"Check(Burn > 14x)"| Page[PagerDuty Alert]
    end

Error Budget Policies

An error budget represents the allowed unreliability: $1 - \text{SLO}$.

• Burn Rate Alerts: Instead of absolute thresholds, alerting is based on the rate of budget consumption. A burn rate of 1 means the budget will be exactly exhausted in the 28-day window.
• Consequence Automation: If the 28-day budget falls below 0, CI/CD pipelines are programmatically frozen to enforce reliability over feature velocity.