Simply put, a data mart is a limited-scale data warehouse and whose data can be obtained by summarizing and selecting data from the data warehouse or with the help of separate extract, transform and load processes from the source data system.
|Basic||The application independent data warehouse.||The specific data mart for the application of the decision support system.|
|Form of data||detailed||summary|
|Use of denormalization||The data is slightly denormalized.||The data is highly denormalized.|
|Data model||From top to bottom||From the bottom to the top|
|Nature||Flexible, data-oriented and long-lasting.||Restrictive, project-oriented and short-lived.|
|Type of scheme used||Constitution of facts||Star and snowflake|
|Ease of construction||Difficult to build||Simple to build|
Definition of Data Warehouse
The term data warehouse indicates a group of time-variant, subject-oriented, nonvolatile and integrated data that assists in the decision process of management. Alternatively, a repository of information collected from multiple sources, stored in a unified scheme, in a single site that allows the integration of a variety of application systems. Once these data are collected, they are stored for a long time, therefore they have a long life and allow access to information historical .
As a result, the data warehouse provides the user with a single interface integrated the data through which the user can easily write decision support questions. The data warehouse helps transform data into information. Designing a data warehouse includes a top-down approach.
It gathers information on topics that affect the whole organization, such as customers, sales, goods, items and, therefore, its range of company level. Generally, the scheme is used constellation of facts, which covers a wide range of topics. A data warehouse is not a static structure and yes evolves continuously.
Definition of Data Mart
A mart date it can be defined as a subset of a data warehouse or a subset of enterprise-level data corresponding to a given set of users. The data warehouse involves several data marts departments is logical which must be persistent in their illustration of the data to ensure the strength of a data warehouse. A data mart is a set of tables that focus on one single activity designed using a bottom-up approach.
Since it scheme of stars is snowflakes guided towards the modeling of single subjects, for this reason that these are commonly used in the data mart. Although, the star pattern is more popular than the snowflake pattern. Depending on the data source, data marts can be classified into two types: data marts independent is independent .
Key differences between data warehouses and data marts
- The application independent data warehouse while the specific data mart for the application of the decision support system.
- The data is stored in a single repository centralized in a data warehouse. By contrast, data mart stores data in a way decentralized in the user area.
- The data warehouse contains a form detailed of data. In contrast, data mart contains data summarized and selected.
- The data in a data warehouse is slightly denormalized while in the case of Data mart highly denormalized.
- The construction of the data warehouse provides a approach from above to the bass . In contrast, the approach is used when building a data mart bottom-up .
- The data warehouse flexible, information oriented and long lasting. In contrast, a data mart restrictive, project oriented and has a shorter existence.
- The fact constellation scheme is usually used for modeling a data warehouse, while in the most popular data mart star scheme.
The data warehouse provides a corporate vision, a single and centralized storage system, an inherent architecture and application independence, while Data mart a subset of a data warehouse that provides a departmental, decentralized storage view. Since the data warehouse is very large and integrated, it presents a high risk of failure and difficulty in its construction. On the other hand, the easy-to-build data mart and the associated lower risk of failure, but the data mart may experience fragmentation.