NOC SLA Implementation for Consumer Trading Platform

Introduction: As the company's production environment becomes more complex, failures in any application can affect overall system availability, prompting the need for an optimized NOC‑SLA system.

C‑end (consumer) overview: The C‑end includes transaction and community scenarios, covering order processing, bidding, inventory, marketing, and recommendation algorithms.

Historical issues: In 2021, C‑end incidents caused order drops, page white‑screens, bidding errors, and other problems, highlighting low alarm discovery (42%) and slow NOC response (3‑15 min).

SLA rationale: To improve alarm quality and response speed, a three‑level SLA (P0 3 min, P1 5 min, P2 15 min) was defined, along with damage‑type categories and business‑view classifications.

Baseline and intelligent baseline: Baselines are set based on historical traffic peaks; an intelligent baseline using Facebook Prophet predicts future metrics and mitigates holiday‑induced spikes.

Alert optimization: Consolidated alerts across application, business, and infrastructure layers; introduced clear rule titles, responsible owners, and damage‑type tags; reduced false‑positives and detection latency.

SLA workflow: Alerts are configured, SLA requests submitted, reviewed by NOC, and, if approved, linked to the SOS emergency system for rapid escalation.

Results: Alert frequency increased (1 min → 10 s collection, 2 min → 20 s rule checks), detection accuracy improved, and incident handling time reduced to under 5 min for critical issues.

Conclusion: The NOC‑SLA framework has enhanced monitoring, reduced noise, and accelerated response, but continuous “freshening” of baselines and rules is required for sustainable reliability.