From Alarm Storms to Proactive Immunity: Geely Auto’s Intelligent Operations Journey

Geely Auto’s rapid business growth and integration of multiple systems across clouds and data centers created exponential complexity, leading to frequent alarm storms where hundreds of alerts appeared simultaneously, making manual root‑cause identification take hours—unacceptable for real‑time automotive services.

To break the limits of traditional operations, Geely chose Alibaba Cloud’s STAROps platform and defined three concrete steps:

Unify Data : Using CloudMonitor 2.0, they aggregated scattered logs, traces, and metrics from dozens of systems into a single storage, enabling consistent analysis and visualization.

Understand Data : Leveraging the UModel (Unified Model) capability, they built an open, extensible model that automatically mapped cloud resources and allowed Geely to inject its own topology and business models, so the AI could “read” relationships between assets.

Continuous Co‑building : Together with STAROps, Geely iteratively refined data preparation and analysis, ensuring the system could answer operational questions accurately and support automated root‑cause diagnosis and emergency recovery.

Key observations emerged:

Without high‑quality architectural assets, intelligent operations cannot succeed; data quality and unified topology are prerequisites.

Introducing AI initially faced resistance—developers preferred manual log inspection, fearing AI hallucinations—but the “true‑smell” effect quickly showed AI could filter noise and suggest remediation within seconds.

In test environments, AI‑driven root‑cause suggestions reduced collaboration overhead, and in production the platform sustained a peak of 50 000 QPS, a standout achievement in the automotive industry.

The role of the ops team shifted from passive ticket handlers to proactive platform operators, focusing on architecture planning, process design, and governance. Geely also envisions embedding STAROps capabilities into developers’ IDEs and CI/CD pipelines, turning intelligent operations into a shared infrastructure for all engineers.

Overall, the collaboration demonstrates that intelligent operations, when built on unified data, open modeling, and continuous partnership, can eliminate operational anxiety, accelerate issue resolution, and enable organizations to move from “people find problems” to “data + AI drive value”.