Based on the process of inflow and outflow in the big data platform architecture, it can be divided into three layers - raw data layer, data warehouse, and data application layer.

1. The raw data layer, also called the ODS (Operational Data Store) layer, is generally obtained from the underlying log data, the business online library, and other sources. Data warehouse data from the ODS layer of data after ETL (ExtractExtra, TransformTransfer, LoadLoad) processing.

2, the main function of the data warehouse is based on the ODS layer data, through the logical processing of the output of the data warehouse theme table. Data warehouse is subdivided into the base layer, theme layer and data mart. The characteristics of the ODS layer are more focused on the query, the variability is large;? Data warehouses are usually at the enterprise level and are used to solve timely, ad hoc problems;? Data marts are more oriented to solving business-specific problems, partly using dimensional models.

3. The data application layer is mainly used to process data from consumer data warehouses.