Course contentsShow
← Software Engineering
Lesson 1178 of 1,93335. Monitoring and AlertingFree lesson

Setting Realistic and Achievable SLOs

Balance customer expectations, system capabilities, and cost when defining SLO thresholds and measurement windows.

Setting Realistic and Achievable SLOs

You'll learn how to define SLO thresholds and measurement windows that balance what customers need, what your system can deliver, and what you can afford to maintain.

The Balancing Act

Setting an SLO isn't just picking a number that sounds good. It's a three-way negotiation between competing forces:

  1. Customer expectations – What level of reliability do users actually need to trust your service?
  2. System capabilities – What can your architecture realistically deliver given current technology and design?
  3. Cost constraints – What level of reliability can you afford to build and maintain?

Think of it like choosing a car: customers might want Ferrari performance, but you need to balance that against what your engineering team can build (maybe a reliable sedan) and what your budget allows (perhaps a well-maintained hatchback).

Practical Guidelines

Start conservative. It's easier to tighten an SLO later than to loosen one that customers depend on. Begin with what you confidently achieve today, not aspirational perfection.

Match the measurement window to reality. A 99.9% availability SLO over 30 days allows ~43 minutes of downtime. Over 90 days, that's ~2 hours. Longer windows smooth out blips but delay feedback.

Consider customer impact, not system perfection. If brief API hiccups don't harm users because they retry automatically, don't set an SLO so strict that it burns your error budget on invisible events.

Factor in operational cost. That extra nine (99.9% → 99.99%) might require doubling infrastructure, on-call staff, and complexity. Is the business case there?

Key Takeaway: Realistic SLOs emerge from honest assessment of current capabilities, explicit tradeoffs between reliability and cost, and deep understanding of what customers actually need—not what sounds impressive.