locked
Is it incorrect to view MDS as a type of Datawarehouse? RRS feed

  • Question

  • We are figuring what type of data constitutes master data for our organization. At present we are dealing with just one entity - Person.
    A person can be an applicant when we first become aware of them. After the application process they can become nothing (they don't progress further), or they can become an employee, student, intern etc).

    So now I have a person entity.
    I have a related PersonType entity.
    It would also be good to store the duration a person spent as a particular PersonType which is probably an SCD.

    Besides fact data, this is all looking very much like a data warehouse, and honestly, something I could much easier without trying to shoehorn it into MDS.

    Am I taking a wrong approach here? Do I need to look at the problem in a different way?

    Wednesday, March 4, 2015 2:54 PM

Answers

  • MDS not a data warehouse. It is a Master Data tool and a Master Data Repository. You would use MDS to modify the Person data, and to provide an authoritative data source for the current Person entities. 

    A Data Warehouse (which could be on the same SQL instance), would hold any fact data, and any historical (SCD) data.  Typically your master data entities would be used to load Data Warehouse dimensions.  And the Data Warehouse dimensions might use an SCD pattern for storing historical data.

    David


    David http://blogs.msdn.com/b/dbrowne/

    • Marked as answer by winstonSmith Wednesday, March 4, 2015 5:23 PM
    Wednesday, March 4, 2015 3:13 PM

All replies

  • MDS not a data warehouse. It is a Master Data tool and a Master Data Repository. You would use MDS to modify the Person data, and to provide an authoritative data source for the current Person entities. 

    A Data Warehouse (which could be on the same SQL instance), would hold any fact data, and any historical (SCD) data.  Typically your master data entities would be used to load Data Warehouse dimensions.  And the Data Warehouse dimensions might use an SCD pattern for storing historical data.

    David


    David http://blogs.msdn.com/b/dbrowne/

    • Marked as answer by winstonSmith Wednesday, March 4, 2015 5:23 PM
    Wednesday, March 4, 2015 3:13 PM
  • That makes sense. can I provide a 2nd scenario and ask how it would be best handled:

    SourceSystem has Persons of type Applicant and test scores for each applicant.
    The persons would be pushed into MDS but MDS is not a place for scores (IMO).

    So any system downstream that wants to look at Applicants and scores will have to pull:
    1. the applicant data from MDS.
    2. Once they have the Applicant data, they then pull the score data from the source system (or DWH if present).

    is that the correct way to approach that problem?

    Wednesday, March 4, 2015 3:34 PM
  • > is that the correct way to approach that problem?

    No.  The Data Warehouse should have a copy of the Master Data, extracted from MDS in addition to the historical transaction data. 

    The Master Data Hub makes the clean, correct, validated Master Data available to any system that needs it.   But often the downstream systems need a copy of the Master Data, not just ad-hoc access. 

    For instance an Order-Entry system needs a copy of the Customer data to perform searches and provide referential integrity; not just runtime customer lookup.  Similarly a Data Warehouse needs to be able to perform searches of transaction data by Customer attributes.  It's just not practical to hit the Master Data Hub to find all the customers California, and then go hit the data warehouse to run a report for those 500 customers.

    David


    David http://blogs.msdn.com/b/dbrowne/


    Wednesday, March 4, 2015 5:50 PM