Methodology and Practice of Onedata Data Warehouse Construction

Overview – The article shares the methodology of constructing an Onedata data warehouse, outlining its main components: methodology system, data modeling workflow, and practical case studies.

1. Data Warehouse Basics – Defines a data warehouse using a real‑world food processing scenario, explains the need for data aggregation, layering, and the transformation from raw data to business‑ready insights.

2. Data Warehouse Methodologies – Introduces the two major schools of thought: Ralph Kimball’s dimensional modeling (fast, low‑cost, but may cause redundancy) and Bill Inmon’s CIF architecture (enterprise‑wide, high‑cost, longer cycle). The CIF model, proposed in 2003, integrates both approaches.

3. Onedata System – Originated by Alibaba, Onedata aims for global data consistency, unified services, and avoidance of duplicate construction. Its three core features are OneID (standardization), OneModel (entity unification), and OneService (service unification).

4. Data Modeling Process – The modeling workflow includes:

Design basis derived from enterprise and business architecture, industry best practices, and application needs.

Model hierarchy division (application layer, aggregation layer, foundational layer) using star/snowflake schemas or third normal form.

Design principles covering unified sharing, standardization, traceability, stability, and scalability.

5. Data Warehouse Design Steps – Requirement analysis, source analysis, gap analysis, conceptual/logic/physical modeling, implementation, validation, and optimization.

6. Case Studies –

Baidu Advertising EDW : Top‑down business architecture mapping to data entities, resulting in ten thematic domains.

Huawei Data Foundation : Emphasizes data lake ingestion, six data standards (owner, standards, sensitivity, source, quality, metadata).

Bank Data Warehouse Models : Comparison of IBM BDM and Teradata FS‑LDM thematic domains, highlighting top‑down design versus bottom‑up approaches.

7. Conclusions – Successful Onedata construction requires top‑level design derived from business architecture, layered data models for stability and scalability, and strict data modeling standards and service interfaces.