By: Nai Biao Zhou || Related Tips:More >Database Design
Problem One technology principle recommended by the Open Group Architecture Framework (TOGAF) is to control technical diversity [1]. To apply this principle, a software development team wants to create a data warehouse with the Microsoft toolset. The team has taken Kimball’s advice [2,3] and considered using the star schema as the data warehouse schema. Furthermore, facts and dimensions have been identified and documented. After team members have pored over Kimball’s other book [4], the team is ready to build a DW/BI system. Due to lack of experience on data modeling, the team wants to know how to manipulate Microsoft tools to create and maintain a star schema data model. SolutionTo practice creating a star schema data model from scratch, we first reviewed some data model concepts and attested that the SQL Server Management Studio (SSMS) has the capacity for data modeling. Then, we created a database through the SSMS, and this allowed us to produce conceptual and logical data models. After looking into the definition of the physical data model, I have made a bold statement that a database project, provided through SQL Server Data Tools (SSDT), could describe a physical data model. Therefore, we generated a physical data model by importing the logical data model, which lives in a database, into a database project in Visual Studio 2017. The database project, containing the object definitions and deployment scripts, fully integrates with source code control software. We committed the database project into the Team Foundation Version Control (TFVC). In the end, through the TFCV, we were able to generate the conceptual and logical model diagrams and an actual database. This guarantees the data model consistency.
The solution was tested with SQL Server Management Studio V17.4 and Microsoft Visual Studio Community 2017 on windows 10 Home 10.0 <X64>, and Visual Studio Team Services (VSTS) [5] in the cloud. The DBMS is Microsoft SQL Server 2017 Enterprise Edition (64-bit). The transactional database schema was retrieved from the AdventureWorks2017 [6]. Create Data Models Regarding data models, there are some variations in the way they are defined. We adopted the fundamental concepts and definitions that are introduced in the book [7]. There are three primary types of data models; conceptual, logical, and physical. Most data modeling tools, listed in [8], accept this classification. However, [8] does not clarify that the SSMS support any type of data models. Due to characteristics of star schema, SSMS, like other tools in [8], fully meets star schema data model development requirements. I would like to point out that some tools in [8] do not support conceptual model. In a data warehousing project, sometimes the conceptual data model and the logical data model are considered as a single deliverable [9]. We usually create a conceptual data model first, then work on a logical data model. After approval of the logical data model, we produce a physical data model. We are going to use dimensions and facts shown in Table 1 to demonstrate the modeling process. Dimensions Facts (1) Date dimension(2) Individual customer dimension
(3) Store dimension
(4) Territory dimension
(5) Sales order number (degenerate dimension)
(6) Employee dimension
(7) Product dimension
(8) Special offer dimension Granularity: by each sales order
(1) Sale subtotal
(2) Tax amount
(3) Shipping cost Granularity: by every product on every order
(1) Quantity ordered per product
(2) Product subtotal
(3) Discount amount
Table 1 Dimensions and Facts in AdventureWorks Data Warehouse
The Conceptual Data ModelA conceptual data model is used to describe entities and their relationships. High-level business users, such as executive managers, can comprehend the model diagram. Nowadays, computer technologies have been widely used. Non-technical users can understand some technology conventions; for instance, the compound noun, Individual Customer, is written as “Individual_Customer”. On the other hand, technology providers work to remove some limitations on naming. The maximum character length of an object in SQL Server 2017 is 128, and SQL Server also allow space in an object name. “Individual Customer” or “Add a New Individual Customer” is a valid object name in SQL Server. This enables business users and technical experts to share a common, rigorous language. Therefore, entity names used in the conceptual model can be used in other two types of data models.
Let’s declare some naming conventions in this exercise. We add a prefix “dim” and a prefix “fact” to dimension table names and fact table names, respectively; and we append a suffix “key” to the entity identifier’s name. We separate two English words with an underscore; for example, the term “dim_individual_customer” represents an individual customer dimension or an individual customer dimension table, and “individual_customer_key” is the identifier’s name of this entity.
A typical data model contains many entities, along with their relationships. To understand a diagram consisting of all these entities and relationships is beyond human’s capacity to process information. Other data modeling tools, for example, “Toad Data Modeler”, can divide a model into multiple workspaces for easier management. One best practice is to create a workspace for each fact table, which splits a complicated model into manageable chunks. Inspired by these Toad practitioners, we are going to create a database diagram in the SSMS for each fact table. Let’s walk through this process step by step.
Step 1: Install Diagram SupportCreate a new database “data_modeling” through the SSMS and select the menu item “Databases -> data_modeling -> Database Diagrams > Install Diagram Support “. The menu item wouldn’t appear if the diagram support feature were installed.
Select the “Yes” in the pop-up window to close the window.
Step 2: Create New Database Diagram
Right-click on the menu item “Database Diagrams” and select the “New Database Diagram” item.
A pop-up window with a title “Add Table” shows up. Click on the “Close” button at the bottom of the new window to close the window. We will use this pop-up window later. If the layout of the SSMS is different from the following screenshot, click on the window’s menu “Window -> Reset Window Layout”.
Step 3: Create User-Defined Data Types
Right-click on the item “Databases -> data_modeling -> Programmability -> Types -> User-Defined Data Types” in the Object Explore panel and select the menu item “New User-Defined Data Type” in the context menu.
A new window appears. Enter a value “udt_surrogate_key” in the name field and select the data type with “int”. Then, click on the “OK” button to create a new data type. As a matter of fact, it is not important to select proper data type at this step.
Follow the preceding procedure describe