In Our environment we have different subject area( lets say 1 to 5). We thought of creating a universe in development environment( subject area 1) and once we feel good on universe, then move universe to production. Then focus on 2, 3 and so on, meaning to do subject area 2 on existing universe in development phase and then move into production.
My question is it good idea to do in phases(subject area wise). Meaning will there be any effect to universe(subject area 1) if we add subject area 2 on existing universe in development phase and then move into production and so on…
Or is it advisable to do all subject areas in development and then move all together into production
From the universe point of view: If you will not use linked universes then it really does not matter because universes will be independent on each other. So if you develop them together and then move them to production or you will develop UNV1, then move it to production and then you move on to a UNV2 - the result will be the same.
From the reports point of view - if your reports will use only a single universe then again, it does not matter. However, if a report will use more universes, then of course you want all the universes to be finished and moved to production before you start building a report.
I missed the point. We are developing Master and derived universes in our environment. Master universe contains core components and derived universe has core components and its own requirements. Meaning First master universe consists fact table and few dimension tables and then in derived universe, we will joins the fact and dimension tables. Lets say we completed Subject area one( ex: sales) and moved to production.
Now second subject area is Marketing. I will work on existing derived universe, meaning will get all fact and dimension tables related to marketing area and add in master universe. Then will add these marketing related tables in the existing derived universe( Subject area one) and do all joins and then export the universe into production.
I hope i am not confusing you. Can we do in phases for master and derived universe???
To try and be clear, are you using linked universes? Or are you adding the Marketing tables to the existing Sales universe, to create kind of a “Sales 2.0: Now with Marketing!” universe? I thought from your first post that you were talking about the latter option, but your second post makes me less sure.
In either case, a phased approach will work just fine. However, I would strongly suggest that you plan as much as possible out ahead of time. Sit down and do up a proper data model that covers both the final product, and that breaks it down into phases. Each time you add tables to the universe, you are likely to end up with some classes that can really benefit from one added field, and some classes that now have redundant fields. You will also have some data that can only be represented as a single field in Sales, that expands into an entire sub-class once Marketing or Manufacturing is introduced. Understanding where these shifts are going to happen ahead of time will save you a ton of work in the long run.
Our plan is to do in phases. Firstly we narrowed down the subject areas as Sales, Marketing and Financing as well as data model. Meaning will focus on one subject area at a time, develop and move into production( both tables and universe). But we want to implement Master and Child concept in our environment to make it simple.
Lets say we created a master universe that consists of fact and dimension tables( Sales) and then created a child universe where we create all joins, context and loops. Once we feel good on universe then we will move into production. So far good.
Now we want to so same with Marketing and Financing areas. Meaning add fact and dimension tables( marketing) in master and then add these tables in the existing derived universe(i.e sales) and do all joins and resolve context. Again if we feel good on the universe then move into production.
My question to experts is a good approach to do in phases using master and derived universe. I guess its would be simple if we create individual universes. Here i want to add all areas in one derived universe in different phases and move into production.
My concern is if i do one subject area at a time, will there be any effect if i do changes in the existing derived universe and move to production. Will there be any effect to first subject area(Sales) which is already in the production??
You question is very confusing . Where did you get this Master, Derived universe terminology from? Are you using linked universes? If yes, PLEASE indicate it in your post.
If you are working on different subject area, create one universe per each subject area. Don’t worry about putting the data together in report, you can do that by linking common dimensions at the report level.
Did this answer your question?
Why would you want a “master” universe of tables and a child universe of joins? It doesn’t make sense.
The only real use of linking universes is if you want to share dimensions, eg: if your Sales and Marketing universes both share a customer dimension. Then it might make sense to have a separate Customer universe with one set of objects and link this into the Sales and Marketing universes.
I do not see the added value in your scenario, which would justify using linked universes.
Linked universes are good if you have one big universe and create derived universe from it by subject area (read: hiding objects),
or if want to take advantage of a Universe Bus System, meaning you have a core universe with all conformed dimension tables and classes/objects, which you link into derived universes.
My bad. I misread the concept here. I am still trying to understand the types of linked universe, i ,e Kernel , Master and Component approach. I am using the concept of linked universe.
Are we good if we do in phases by using linked universe??
I would really appreciate if someone tell me when we use different types of linked universe