| Author |
Topic |
|
Ardvisoor
Starting Member
3 Posts |
 Posted - 2006-01-15 : 01:34:08
|
| Hi, i Studied Both Relational and Multidimentional model for Data Warehousing,but i don't know which is the best to do? can u Please Guide me?ThxMaryam |
|
|
blindman
Master Smack Fu Yak Hacker
2365 Posts |
Posted - 2006-01-15 : 11:42:27
|
| You mean, Inmon's approach vs. Kimball's?IMHO, Inmon's methodology is the best approach for true data warehouseing. Kimball's is a quick and satisfactory approach for data marts, but his emphasis on star schema's and one-style-fits-all design approach overly simplistic and is not appropriate for enterprise data management. |
 |
|
|
rockmoose
SQL Natt Alfen
3279 Posts |
Posted - 2006-01-15 : 16:22:38
|
| The "multidimensional approach" preached by kimball & co. to me is more like a "quick&dirty" approach.Sure it can work, things can be delivered, but it will always, always come back and bite your ass eventually.If you do a (properly indexed) normalized database, based on a logical model that correctly reflects the business, you can't shoot far from the mark.Here are some articles discussing Inmon an Kimball : http://www.b-eye-network.com/authors/index.php?a=3rockmoose |
|
|
|
blindman
Master Smack Fu Yak Hacker
2365 Posts |
Posted - 2006-01-16 : 10:51:28
|
quote:
Originally posted by rockmoose
The "multidimensional approach" preached by kimball & co. to me is more like a "quick&dirty" approach.Sure it can work, things can be delivered, but it will always, always come back and bite your ass eventually.
Glad to see I'm not the only dba that feels this way. In my personal opinion, Bill Inmon sells the concept of data warehousing, while Ralph Kimball sells seminars. |
|
|
|
Michael Valentine Jones
Yak DBA Kernel (pronounced Colonel)
7020 Posts |
Posted - 2006-01-16 : 11:59:51
|
| Since this was posted on the Analysis Services forum, I just thought I would jump in to point out that if you are going to use Analysis Services, you will need to have some kind of Multidimentional schema in a database to be able to load your Analysis Services cube(s).These are not necessarily mutually exclusive choices, either. You can have Datamart schemas that are loaded as downstream feeds from a Data Warehouse. Even if you bypass the full Data Warehouse, you will typically have some form of relational model in your Datamart in order to be able to maintain the Multidimentional schema.I think Kimballs approach is to emphasize a subject area approach for the specific purpose of making reporting simple, and I think it does a good job with data that people want to be able to explore on an ad hoc basis.Like everything else, the best tool for the job depends on the job you are trying to do. I think one of the big mistakes that people make with a data warehouse is to decide to build one without asking the question of what business purpose you are trying to serve. Kind of taking the approach of “If I build it, they will come.” Once these questions are asked and answered, then it becomes a lot easier to see the direction you need to go.CODO ERGO SUM |
|
|
|
blindman
Master Smack Fu Yak Hacker
2365 Posts |
Posted - 2006-01-16 : 13:30:32
|
No arguments, but clarification:quote:
Originally posted by Michael Valentine Jones...You can have Datamart schemas that are loaded as downstream feeds from a Data Warehouse.
This is what Inmon advises. Spinning off smaller subject oriented data marts for focused analysis, while maintaining the data warehouse as the point of record.quote:
Originally posted by Michael Valentine Jones...I think Kimballs approach is to emphasize a subject area approach for the specific purpose of making reporting simple...
Again, making it fine for a datamart, but bypassing what I consider necessary for a true Enterprise-wide multi-disciplinary data warehouse. |
|
|
|
rockmoose
SQL Natt Alfen
3279 Posts |
Posted - 2006-01-16 : 16:55:41
|
quote:
I just thought I would jump in to point out that if you are going to use Analysis Services, you will need to have some kind of Multidimentional schema in a database to be able to load your Analysis Services cube(s).
Multidimentional schema - ?A n-ary relation has n dimensions.I don't like that term, it's just a database schema, you still need a solid logical model as foundation.All schemas are "multidimensional", and I often see that term in association with "denormalization".Denormalizing a schema does not make it more "dimensional".It is not a prerequisite to denormalize anything at all if you want to load Analysis Server cubes.In fact it will work just as well on a fully normalized structure, I imagine that is a common misconception.However, for loading AS cubes I do find it beneficial (and often necessary) to massage the data before loading it into cubes.By that I mean adding formatting columns, filtering criteria, precalculations, some denormalization etc...in order to make the source data work well with the tool (AS) and the desired user experience.All in all, I am personally more in line with Inmon approach of a fully normalized datawarehouse from which you may retrieve the data into subject areas (marts).And/Or retrieve the data in whatever format is needed for the receiving end (ftp/reports/exports...).A central master record for all data is the key, from that you may do anything.rockmoose |
|
|
|
rmason
Starting Member
12 Posts |
Posted - 2006-02-09 : 12:26:31
|
| All of the comments thus far have very good arguments behind them and I think that this type of debate will keep coming up for a long time to come. Of couse a full enterprise wide data warehouse is the ideal and is something that all BI developers should be aiming for in their developments. However, the real world can be a cruel and cold place and businesses don't like pouring vast amounts of money into projects with nothing in return apart from the reassurance that everything is on target. This in my opinion, is where Kimball's approach is more practical as you can start delivering results to the business far earlier on.Kimball doesn't actively recommend not building a fully blown data warehouse but instead advocates a backfill approach. Of course this may mean redesigning some aspects of the warehouse over time to cater for unforeseen requirements but this is far more palitable if the project is already delivering results.In summary the approach taken really depends on the scope and budget behind a project and how quickly the business expect to see results.Rob. |
|
|
|
|
WHICH Model is Better For WareHousing?