Understanding SLA, SLO, and SLI: Concepts, Practices, and Alert Governance for High‑Traffic Events

The article begins by introducing common confusion around service‑level agreements (SLA), service‑level objectives (SLO), and service‑level indicators (SLI), especially during high‑traffic periods like the 11.11 promotion, and promises practical guidance based on real‑world experience.

It defines SLI as a quantifiable metric of service quality (e.g., availability, TP99 latency, throughput, persistence), SLO as the target value or range for a specific SLI, and SLA as the contractual agreement that specifies consequences when an SLO is not met.

The benefits of SLOs are described for both clients (predictable quality) and providers (clear quality standards, cost/benefit trade‑offs, risk control, faster incident response).

A service‑grading model is presented, classifying applications and interfaces into levels (0‑3) based on business impact, and explaining how grading drives alert policies, reliability requirements, and operational controls.

Common pitfalls such as inconsistent grading, inaccurate availability reporting, and noisy alerts are discussed, along with solutions like unified grading, traceability from user perspective, and code‑level availability monitoring.

Practical SLI selection guidance recommends focusing on 3‑5 representative metrics (e.g., availability, request volume, TP99) rather than tracking every possible metric.

SLO‑driven alert governance is detailed: alerts should be triggered only when error‑budget consumption exceeds thresholds, with separate critical and warning thresholds for availability and latency, and recommendations for alert channels (e.g., DingTalk, email, voice).

Specific alert configurations are provided for API, message‑queue, and scheduled‑task monitoring, illustrating how to set thresholds, combine metrics, and avoid unnecessary noise.

The article answers three common questions: the relationship between SLA, TP99, and timeout settings; how to calculate departmental availability from individual services; and how to compute overall availability when a subset of cloud hosts fails.

It concludes with a checklist for using SLA/SLO to prepare for large promotions: defining clear service goals, capacity planning, full‑link stress testing, monitoring and alerting, priority management, cross‑team collaboration, and post‑event continuous improvement.

Appendices include a sample SLO document, references to national standards and major cloud provider SLA documents, and a list of further reading.