With the advent of the era of big data, data warehouse, as the core component of data processing, has attracted more and more attention. Whether the model in the data warehouse is reasonable or not directly affects the accuracy and efficiency of data analysis and decision-making. Therefore, understanding the models in a data warehouse is an essential skill for every data warehouse engineer and data analyst. In this article, we will focus on three types of models in a data warehouse: the bus schema approach, the dimensional model, and the fact model.
1. Bus architecture method (classic structure of BW bus)
The bus architecture method is a method that divides the data of the enterprise information system into four levels: functional domain, sub-functional domain, business process, and information processing process. Each information processing process includes three parts: input, processing and output. In the data warehouse, we can use the bus architecture method to divide enterprise data according to business domains, making data processing more efficient and orderly.
2. Dimensional model (analysis-oriented database design)
Dimensional model refers to a data model that divides data according to dimensions and measures according to analysis requirements. Among them, dimension refers to various angles describing data, such as time, location, product, etc.; measurement refers to the numerical index of data, such as sales, profit, etc. The design principles of the dimensional model include: each dimension is related to at least one measure, otherwise it can be considered to be split into two dimensions; the dimension cannot have an inheritance relationship, otherwise the top-level dimension can be split into multiple dimensions; Use calculation results as attributes on dimensions, because this will lead to data non-traceability.
3. Fact model (measuring business performance)
The fact model refers to a data model that describes the business performance of an enterprise. It usually contains multiple fact tables, each describing a specific business performance metric, such as sales, visits, and so on. Each fact table is related to at least one dimension table, and the data of the fact table is described through the attributes of the dimension table. The design principles of the fact model include: each fact table should have a timestamp, which is used to record the time of data; each fact table should have a measure value, which is used to describe the value of the data; the measure value of the fact table should be as far as possible Use raw data to ensure data accuracy.
Summary: The three types of models in the data warehouse are the core technology for building a data warehouse and the basis for data analysis. The bus architecture method can help us divide the enterprise data in an orderly manner, the dimensional model can help us better understand the data, and the fact model can help us measure the business performance of the enterprise. In practical applications, we can choose an appropriate model according to specific business needs to achieve the best data processing and analysis results. Therefore, learning and mastering these models is crucial for every data warehouse engineer and data analyst.
This article is published by mdnice multi-platform