Roll-Up or Drill-Down, which has the largest Data Warehouse detailing?



  • Context

    Studying about data warehouse(DW) came to me with the content of OLAP server types and the types of operations that can be done, within that context I was confused with the concept of detailing about two operations, are they called Roll-Up and Drill-Down

    In the book, the author states that the Drill-Down operation has greater detailing but the given example did not make me understand well.

    Exemplification of the Book.

    Chapter 29 - Overview of Data Warehousing and OLAP. Below is how operations are visually exemplified by the author of the book.

    Example Roll-Up

    inserir a descrição da imagem aqui


    Example Drill-Down

    inserir a descrição da imagem aqui

    What the Teradatapoint site affirms

    Since it wasn't very clear the explanation, I went to google and found this inserir a descrição da imagem aqui

    My Understanding

    The content illustrated on the teradatapoint site, made me better understand the subject and if I could summarize or explain in a few words the operations, it would be more or less like this:

    Roll-Up Operation seeks reduce the 3D cube for a leaner analysis, for example: Perform a bimetral analysis of a given product in a given region.


    Drill-Down operation seeks increase the 3D cube, for an analysis with a longer time interval, for example conducting analyses at annual intervals of a given product in a given region.

    I understand by detailing the data, an increasingly smaller delimitation of the scope involved, which indicates a decrease in the size of the cube dimension, but the book states the opposite.

    Doubt

    I was unaware of which of the two operations there is a greater data detailing, the book states that it's Drill-Down and the terapoint site says it's Roll-Up.

    Source: Database Systems 6th edition, authors: Elmasri, Ramez Navathe, Shamkant B. Year: 2011 Cap 29 pg 724 and 725



  • Well, I'm a rookie here, and this is my first answer. As my knowledge of the subject is also not of the deepest, I will try to be as summarized as possible.

    OLAP

    When the database is built in OLAP, it means that it is no longer relational, meaning there is no increase or decrease in the amount of data, they are already in a format whose recovery is usually faster than in a relational database. However, there are OLAP versions, ranging from ROLAP (with a relational motor), MOLAP (with a multidimensional motor) and an HOLAP intermediate (Hybrid). Unfortunately, I've never worked with any hybrids to give more explanations.

    About Drill-Down and Roll-up

    When you access the information from an OLAP database (i.e. Multidimensional) it is possible that it has two or more dimensions. Some of these dimensions may have hierarchies. The higher hierarchies are the aggregated information of the lower hierarchies, for example: One side of the dimension can be built with customer location information. In this case, Brazil would be the highest hierarchy. By applying a Drill-down, this hierarchy can be divided in various ways, such as regions (North, South, Southeast etc.) or states (SP, RJ etc.). How much lower you go down in detailing the information, the more the chance you find cells without measures in the same way that the information becomes detailed.

    As for Roll-Up, you can think about the product dimension as it becomes easier to understand. You may be organizing the database of a retail store with various products. Therefore each product would be the lowest cell of the hierarchy of this dimension. In this way, when applying the aggregation of the sales of a particular product group (Liquifier -> Kitchen Products -> Household Appliances, for example), the complication existing at this point is the fact that the information is not standardized among the products, for example, which will require some form of data transformation.

    I hope I did. Hugs!




Suggested Topics

  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2