SLIs & SLOs

PRISM supports Service Level Indicators (SLIs) and Service Level Objectives (SLOs) following SRE best practices.

SLI (Service Level Indicator)

An SLI defines what is being measured.

Fields

Field	Type	Required	Description
name	string	No	SLI name
description	string	No	SLI description
formula	string	No	Calculation formula

Example

{
  "sli": {
    "name": "Availability",
    "description": "Percentage of successful requests",
    "formula": "successful_requests / total_requests * 100"
  }
}

SLO (Service Level Objective)

An SLO defines the target for an SLI.

Fields

Field	Type	Required	Description
target	string	Yes	Human-readable target
operator	string	No	Comparison operator for machine evaluation
value	number	No	Numeric target value
window	string	No	Measurement window
thresholds	object	No	Additional thresholds

SLO Operators

PRISM supports machine-evaluable SLOs with these operators:

Operator	Constant	Description	Example
gte	>=	Greater than or equal	Availability ≥99.99%
lte	<=	Less than or equal	Latency ≤200ms
gt	>	Greater than	Score >80
lt	<	Less than	Error rate <0.1%
eq	=	Equal to	Target exactly 100

Example with Machine-Evaluable SLO

{
  "slo": {
    "target": ">=99.99%",
    "operator": "gte",
    "value": 99.99,
    "window": "30d"
  }
}

Measurement Windows

Common window values:

Window	Description
7d	7-day rolling window
30d	30-day rolling window
90d	90-day rolling window
monthly	Calendar month
quarterly	Calendar quarter

Programmatic SLO Checking

PRISM provides a MeetsSLO() method for programmatic checking:

metric := prism.Metric{
    Current: 99.95,
    SLO: &prism.SLO{
        Target:   ">=99.99%",
        Operator: prism.SLOOperatorGTE,
        Value:    99.99,
    },
}

if metric.MeetsSLO() {
    fmt.Println("SLO met!")
} else {
    fmt.Println("SLO not met")
}

Operator Behavior

Operator	Current	Value	MeetsSLO()
gte	99.99	99.99	true
gte	99.95	99.99	false
lte	200	250	true
lte	300	250	false
eq	100	100	true

Complete Metric Example

{
  "id": "ops-availability",
  "name": "Service Availability",
  "description": "Percentage of time the service is available",
  "domain": "operations",
  "stage": "runtime",
  "category": "reliability",
  "metricType": "rate",
  "trendDirection": "higher_better",
  "unit": "%",
  "baseline": 99.0,
  "current": 99.95,
  "target": 99.99,
  "sli": {
    "name": "Availability",
    "description": "Successful requests / total requests",
    "formula": "1 - (error_count / total_requests)"
  },
  "slo": {
    "target": ">=99.99%",
    "operator": "gte",
    "value": 99.99,
    "window": "30d"
  },
  "thresholds": {
    "green": 99.95,
    "yellow": 99.9,
    "red": 99.0
  },
  "frameworkMappings": [
    {"framework": "SRE", "reference": "availability-slo"},
    {"framework": "DORA", "reference": "availability"}
  ]
}

Best Practices

1. Set Realistic Targets - SLOs should be achievable but challenging
2. Include Error Budgets - Use thresholds to define acceptable ranges
3. Document Formulas - Include SLI formulas for clarity
4. Use Machine-Evaluable Operators - Enable automated SLO checking
5. Define Measurement Windows - Clarify the evaluation period
6. Map to Frameworks - Reference industry standards (SRE, DORA)