SQL Server Performance

Should I define my dimensions as 'Changing'?

Discussion in 'Analysis Services/Data Warehousing' started by Joozh, Jan 6, 2005.

  1. Joozh New Member

    Hi,

    The approach that I am using is that I drop, create, and then populate all the tables in my DataMart. And after populating my tables, I am doing a full processing of allmy dimensions and then the cube.

    Given this approach and my understanding (so far), I am assuming that I need not define my dimension as changing.

    Can someone kindly confirm if I am using the right approach or whether I still should define my Dimensions as 'Changing Dimension'. If so, kindly help me understand the benefits of doing so.

    Thanks.
  2. thomas New Member

    No, I don't think so.

    Tom Pullen
    DBA, Oxfam GB
  3. satya Moderator

  4. Raulie New Member

    If you are performing a full process then there is no need to define changing dimensions. Refer to Books online "Changing Dimensions".

    Raulie
    Hewlett-Packard Company
  5. Joozh New Member

    Hi Tom, Satya, and Raulie,

    Hope 2005 is off to a great start for all of you and please accept my best wishes for the year...lots of happiness, success, peace of mind, and all that good stuff.

    Thanks for the replies.

    Satya: Thanks for the URL. I am becoming a BIG fan of William Pearson(Bill) - whose article is in the URL. Though he has tons of articles but so far I have only been able to read a few of these. Hopefully this one too will as good as the few that I have read.

    Bill, Tom, Satya:
    Will read BOL too but for now just a thought (a scarry one though). In my current situation I am talking about 1 million records for the fact, and say about 500 each for other dimensions (Customer, Product, etc.). Since my underlying data is not that much therefore I my current approach i.e. Droping, Creating, and then Populatingmy tables is working find and the whole process (starting with the dropping of tables and all the way to the full process of the last cube takes about 3 minutes). COming to my question:

    What if I am asked to develop another one but thistime with milliongs of records for the fact and at least thousands for the dimensions... Maybe the article will help but I can't think of how to handle that??? in theearly daysof my learning I tried with bringing in only the 'newer' records but then faced a lot of difficulty in determining how to separate new records from old if the datamart is updated 2 or 3 times a day.....

    Can you kindly share based on your experience.

    Thanks.

Share This Page