How a Distributed Big‑Data Architecture Turned Million‑Query‑Per‑Day Account Lookups from Minutes to Milliseconds

Amid rapid digital‑economy growth, banks are shifting from "business support" to "technology‑led" models. The account transaction detail query, handling more than one million daily calls, is a critical "digital mirror" of corporate cash flow, but the legacy centralized architecture could not deliver millisecond‑level responses for massive historical data.

Breakthrough: New Distributed Intelligent Query Architecture

The project followed a "business‑driven, technology‑enabled" principle with four goals—performance boost, data integration, experience optimization, and compliant controllability. By separating compute from query, the system achieved high availability, high performance, and strong security.

Technical Path

A distributed micro‑service architecture provides horizontal scaling for high‑concurrency queries. Leveraging the Apache Spark and Hadoop ecosystem together with Huawei’s GaussDB, the solution overcomes relational‑database limits and forms a complete stack of data services, storage, compute, and system resources. The implementation combined top‑down design with agile sprints to ensure smooth rollout.

Core Innovations

Millisecond‑level retrieval of massive offline data: using Spark/Hadoop, multi‑dimensional indexes, intelligent caching, pre‑computation, and fine‑tuned query optimization, the system can precisely fetch data older than ten years.

Modular, unit‑based architecture that separates compute and storage, organized into data‑service, storage, compute, and resource layers.

Lambda processing architecture in the compute layer delivers a dual‑engine model that blends near‑real‑time low‑latency processing with high‑throughput batch handling, providing elastic scaling based on workload.

Performance Leap

Average response time dropped to under 200 ms—a 700% improvement. Query time span expanded from 31 days to ten years, and maximum returned rows grew from thousands to hundreds of thousands, enabling “one‑click aggregation, instant availability” for over 40,000 branches and dramatically shortening customer transaction processing.

Data Integration

Cross‑departmental effort migrated all historical data since 2008, reviving previously siloed “sleeping” data and creating a continuous lifecycle view that eliminates data gaps and enhances financial traceability.

Experience Enhancements

The system supports flexible ordering (ascending/descending), field‑level selection, and adapts to accounting reconciliation, audit tracing, tax filing, and other scenarios. For sensitive use‑cases such as bidding or judicial queries, a dual‑track masking mode hides details at both record and field levels, balancing efficiency with strict compliance.

Compliance and Risk Control

Optimizing the counterpart ledger architecture resolves complex “one‑to‑many” identification challenges in services like social‑security payroll, embedding precise risk‑control checks directly into business flows and markedly improving due‑diligence accuracy.

Operational Management

Standardized receipt management consolidates hundreds of legacy templates into four unified formats, establishing a cross‑channel receipt system with three‑fold uniformity in format, generation rules, and displayed information. Multi‑dimensional filtering and one‑click PDF generation boost business handling and financial reconciliation efficiency.

Future Outlook

Having successfully launched the next‑generation query function, the bank plans to evolve from precise retrieval to intelligent insight by integrating AI and large‑model technologies, turning the system into a proactive analysis partner that offers deeper financial insights and decision support.