SLOs for Different Service Tiers
What you'll learn: How to set appropriate, differentiated SLO targets based on service criticality and customer tier, balancing user expectations with operational cost.
Why Not All Services Are Equal
Not every service requires five-nines availability. A critical payment processing API demands stricter SLOs than an internal reporting dashboard used once a month. Similarly, premium customers may expect better performance than free-tier users. Setting differentiated SLOs helps you allocate engineering effort where it matters most.
Think of it like shipping services: overnight express requires tighter delivery guarantees (and costs more) than standard ground shipping. Both work, but the SLOs differ.
Service Criticality Tiers
Critical services typically need:
- Availability: 99.95% or higher
- Latency (p99): <100ms
- Tight error budgets with aggressive burn-rate alerts
Standard services might use:
- Availability: 99.5%
- Latency (p99): <500ms
- More relaxed monitoring thresholds
Non-critical services could accept:
- Availability: 99% or lower
- Latency: measured at p95 instead of p99
- Larger error budgets
Customer Tier Differentiation
Premium customers might receive SLAs backed by SLOs like 99.9% availability, while free-tier users operate under "best effort" with no formal guarantee. Your internal SLI specifications and error budget calculations should reflect these tiers—premium traffic might have separate SLIs tracked independently.
This doesn't mean degrading service intentionally; it means prioritizing incident response and optimization work based on business impact.
Key Takeaway: Differentiated SLOs align engineering investment with business priorities—set stricter targets for critical services and premium tiers, while allowing looser guarantees elsewhere to optimize cost and effort.