Design and Practice of the 58 Agile BI System (Starfire)

Starfire is a next‑generation SaaS‑enabled data analysis and visualization product developed by 58.com, offering agile BI, visual reports, dashboards, and data‑screen capabilities to support end‑to‑end data insight workflows.

The platform consists of five major modules—analysis workbench, operation workbench, open services, cockpit, and StarMap—allowing users to create datasets, design dashboards, and interact with components such as filters, charts, and export functions.

Its architecture is built on micro‑services registered in Nacos, with a Spring Cloud Gateway handling routing, gray releases, and logging. Backend services include BI services, Starfire ecosystem services, and data scheduling via the StarRiver system, while data storage uses MySQL for system data, Redis for caching, and ClickHouse as an MPP acceleration layer.

Starfire implements a comprehensive RBAC‑based permission system covering functional and data permissions, with fine‑grained row and column controls; permission checks are performed using binary bit operations, e.g., select operation&2 from user_permission where user_id=1 and resource_id=1 .

The high‑performance query engine separates fast‑compute requests from regular queries, routing them through dialect adapters, SQL optimization, and either direct JDBC access or MPP extraction, and supports common BI calculations such as period‑over‑period, ratios, and cumulative metrics.

MPP cache acceleration is triggered for Hive tables, non‑SQL sources (CSV/Excel), and internal document systems, with configurable extraction switches and manual extraction options.

Starfire also provides a visualization resource library (StarMap) that combines a community‑driven chart library and the Wcharts component set, enabling users to create and share custom front‑end components.

Developer services include OpenAPI for resource listing, permission management, and user management, as well as embed‑service APIs and iframe SDKs for seamless integration of reports into external applications.

Future plans aim to expand data source support (APIs, streaming, non‑SQL), enhance lightweight ETL capabilities, introduce richer visual components, and add intelligent features such as alerts and attribution analysis.