AIOps in Banking: Veteran’s Secrets to Smarter Operations

Interview Overview

Yuan Chunliang, a systems analyst at the Bank of China Software Center, reflects on more than 20 years of experience, from his early software development career after graduating in 1999 to his current focus on operations and AIOps.

AIOps Journey at the Bank

Since the 2014 integration of the maintenance department into the software center, the bank has pursued tighter development‑operations collaboration. In 2016 Yuan moved from development to maintenance, witnessing the rise of DevOps and AIOps. The 2017 X86 migration of the core banking system introduced a distributed architecture with hundreds of virtual nodes, creating a need for efficient, safe operational tools.

Early automation efforts included a Dubbo‑based transaction statistics and monitoring tool and a QREP‑based data latency detection tool.

No‑Threshold Intelligent Monitoring

The bank developed a “no‑threshold” monitoring system that automatically judges normalcy based on historical data, calculates risk probabilities for each metric, adapts to dynamic system changes, and operates without intrusive instrumentation.

The core technologies are three models: time‑series prediction, risk‑identification, and adaptive alarm models. These models work together to provide proactive alerts.

Monitoring Coverage

System‑level metrics such as CPU, memory, database connections, MQ depth, and disk space.

Application‑level metrics including TPS, transaction response time, and success rate.

Business‑level metrics such as customer growth and foreign‑exchange rate fluctuations.

An example highlighted a disk‑space alert caused by excessive Zookeeper log writes; after identifying the bug, the team upgraded Zookeeper, eliminating the risk.

Multi‑Indicator Intelligent Monitoring Tools

Beyond single‑metric monitoring, the bank built tools for multi‑dimensional monitoring, such as an online system joint monitor and a neural‑network‑based MQ queue depth predictor, demonstrating strong predictive performance in complex scenarios.

Full‑Process Intelligent Ticket System

To address growing ticket volume, the team created an AI‑driven ticket assistance system that classifies tickets, suggests handling procedures, and automates routing and reminders, significantly improving processing efficiency.

Future Directions

The bank plans to extend AIOps to cloud‑native environments, establish unified data sources, and adopt SRE‑style proactive operations, ensuring resilience amid large‑scale migrations.

Advice for Young Operations Professionals

Yuan emphasizes solid technical foundations, responsibility, system‑wide thinking, and continuous automation while being aware of associated risks.

Conference Highlights

Yuan presented these experiences at the 2019 GOPS Global Operations Conference in Shanghai, encouraging peers to share knowledge and celebrate the evolving role of operations.