How Bull Group Boosted Observability by Migrating from SkyWalking to Alibaba Cloud ARMS

Background and Architecture Upgrade

As a leading domestic electrical equipment company, Bull Group continuously pursues safe, intelligent, and reliable power solutions. With digital transformation, its applications have evolved from monolithic to micro‑service, cloud‑native architectures, creating complex topologies, frequent service calls, and massive runtime data that demand a robust observability platform.

From SkyWalking to ARMS

The team initially used the open‑source APM tool SkyWalking, which met early tracing needs but later showed performance bottlenecks as micro‑service counts grew. To shift from passive response to proactive governance, Bull Group chose Alibaba Cloud Application Real‑Time Monitoring Service (ARMS) for a full‑stack observability solution that integrates tracing, logging, metrics, and intelligent alerts, especially valuable in AI‑IoT scenarios.

Overall Observation Chain

User local gateway → voice input → ASR (speech recognition) → MultiAgent → IoT command execution → reply text generation → TTS (speech synthesis) → device response.

Technical Selection Criteria

Key dimensions compared included integration complexity, query capability, deep‑analysis ability, and probe performance impact. Commercial solutions offer standardized probes, auto‑discovery, and visual configuration, reducing deployment effort, while open‑source tools require extensive custom integration and can strain resources in heterogeneous environments.

Migration Process

One‑click integration: Enable link tracing in the cloud console to view call chains instantly.

Automatic instrumentation: ARMS provides self‑developed probes for Java, Go, Python, improving insertion quality, performance, and stability without major code changes.

OpenTelemetry support: Standard protocol ensures complete call‑chain reconstruction, topology, and dependency analysis.

Post‑Migration Benefits

After migration, ARMS covers hundreds of micro‑service nodes with high‑performance tracing, real‑time metric analysis, intelligent anomaly detection, and unified alerting. Mean Time To Repair (MTTR) dropped by over 60%, and operational efficiency and system stability significantly improved.

Cross‑Domain Value (AI & IoT)

ARMS enabled end‑to‑end tracing of LLM inference, IoT command execution, and voice synthesis, pinpointing latency bottlenecks (e.g., ASR accounting for 65% of delay) and optimizing large‑model response times. Real‑time monitoring of model APIs, token generation speed, and TTS failure rates allowed proactive quality assurance and resource optimization.

Future Plans

Bull Group aims to integrate AI‑driven root‑cause analysis, fine‑grained cost management via dynamic sampling, hybrid‑cloud unified observability, and deeper model‑level metrics such as GPU utilization and KV‑cache hit rates, turning observability into a strategic foundation for digital innovation.