Este sitio utiliza cookies. Si continúa navegando en este sitio, acepta nuestro uso de cookies. Lea nuestra política de privacidad>

Búsqueda
  • Bank of Communications Establishes Benchmark with Resilient AI-powered Storage O&M

    Bank of Communications Establishes Benchmark with Resilient AI-powered Storage O&M

As the intelligent era unfolds, data centers in the financial industry are experiencing exponential growth and becoming increasingly complex. How can we shift from alarm-driven, reactive response to proactive prevention? How can we efficiently identify system risks from massive volumes of logs?

Our goal is to free O&M engineers from tedious and labor-intensive work, so that time-consuming, needle-in-a-haystack troubleshooting becomes a thing of the past. This is far more than a tool upgrade; it is a breakthrough that transforms the O&M paradigm.

As one of China's biggest banks, the Bank of Communications provides mission-critical services essential to the economy and people's livelihoods. Ensuring uninterrupted services and zero data loss is the top priority for its data center team.

However, conventional O&M dependent on alarms and metric analysis struggle to identify risks early enough for proactive prevention. In an emergency, troubleshooting relies heavily on the experience of individual engineers, which makes it difficult to isolate faults quickly. Against this backdrop, a key challenge emerges: how can technologies such as AI agents and knowledge bases be leveraged to enhance the reliability and efficiency of IT system O&M?

The Bank of Communications' data center team and Huawei DME storage data management R&D team, introduced an innovative 1+1+N multi-agent intelligent O&M architecture. This architecture enables each AI agent to perform its role while collaborating seamlessly with other agents, ensuring the reliability and stability of financial IT infrastructure. 1+1+N denotes:
• One brain: dynamic task decision-making engine (knowledge graph + RAG)
• One hub: workflow orchestration engine enabling concurrent scheduling of many tasks
• N agents: all-purpose agents providing collaborative compute, storage, and infrastructure functions

On October 26, 2025, the solution won first prize in the AI+ Application and Skills Competition of the AI Pioneer Cup, setting a benchmark of integration and innovation in AI + Financial O&M.

Certificate for first prize of the AI Pioneer Cup

Certificate for first prize of the AI Pioneer Cup

Conversational AI for Storage O&M

In the past, engineers needed to log in to multiple systems during troubleshooting, relying on personal experience to comb through massive volumes of logs in search of anomalies. Today, at the Bank of Communications' data center, O&M engineers only need to say:

"Analyze the cause of VM A's CPU performance issue."

Bank of Communications Establishes Benchmark with Resilient AI-powered Storage O&M

The DataMaster O&M agent can understand human intent, decompose tasks automatically, invoke APIs and data, generate analysis reports, and provide optimization recommendations. O&M personnel no longer need to perform complex operations manually. Underlying this transformation is the integration of natural language-to-API (NL2API) technology and the RAG knowledge engine, enabling the large AI model to understand O&M, reason intelligently, and execute tasks efficiently.

In practical deployments, the solution has demonstrated an industry-leading accuracy of over 90% for single-turn Q&A accuracy, and over 85% for multi-turn dialogue comprehension and integration.

AI Dual-Mind: Fast Thinking + Deep Thinking for Fault Location in Minutes

Faced with tens of billions of logs every day, conventional rule engines cannot cope with complex issues, while pure large AI models suffer from a high response latency. The dual-path diagnostic mechanism of DataMaster resolves these issues with fast thinking and deep thinking.
• Fast thinking: Rule-based matching delivers responses in seconds for common issues.
• Deep thinking: By integrating statistical features with semantic embeddings and combining LLMs with fault knowledge graphs, the system enables deep root-cause analysis.
This system pinpoints anomalies in minutes—an unprecedented speed and precision.

The intelligent O&M architecture jointly developed by Bank of Communications and Huawei is transforming the storage O&M paradigm in the financial industry. By forming a complete closed loop of perception, decision-making, and execution, DataMaster not only improves O&M efficiency, but provides a path for digital transformation through human-AI collaboration. As AI becomes a veritable intelligent assistant, we move one step closer to the ultimate goal of zero downtime.

The DataMaster AI O&M agent has made storage operations effortless and seamless through the innovative use of AI dialogues.

Arriba