SSAS 'Fact Dimension' v 'Regular Dimension' your views
2 posters
Page 1 of 1
SSAS 'Fact Dimension' v 'Regular Dimension' your views
Hi,
I have been a user of this forum for a while but haven't registered until now. First of all, hello fellow BI developers :-)
I have put a similar post on the MSDN BI Section but thought I would post here as I could get different views/answers from other BI developers. I have general design question regarding the use of a 'Fact Dimension' table relationships as can be specified in SSAS 2005+ versus seperate Fact and Dimension tables (i.e. Regular Dimensions).
I'm finding that a fair amount of my transaction data has several attributes specific to the transaction only (degenerate dimensions 'DD' I beleive is the term). Is it ever worth spliting this data into two tables (Fact and Dimension) when dealing with 'DD' or is this only necessary for regular dimensions which span multiple facts? I get the impression there are performance gains for doing this when the DD has large string values or does SSAS take care of this when you speficy the measures and dimensions from the one table.
In summary, when do people choose to use one design versus the other?
Thanks
I have been a user of this forum for a while but haven't registered until now. First of all, hello fellow BI developers :-)
I have put a similar post on the MSDN BI Section but thought I would post here as I could get different views/answers from other BI developers. I have general design question regarding the use of a 'Fact Dimension' table relationships as can be specified in SSAS 2005+ versus seperate Fact and Dimension tables (i.e. Regular Dimensions).
I'm finding that a fair amount of my transaction data has several attributes specific to the transaction only (degenerate dimensions 'DD' I beleive is the term). Is it ever worth spliting this data into two tables (Fact and Dimension) when dealing with 'DD' or is this only necessary for regular dimensions which span multiple facts? I get the impression there are performance gains for doing this when the DD has large string values or does SSAS take care of this when you speficy the measures and dimensions from the one table.
In summary, when do people choose to use one design versus the other?
Thanks
Ham09- Posts : 8
Join date : 2011-07-26
Location : United Kingdom
Re: SSAS 'Fact Dimension' v 'Regular Dimension' your views
Maybe I should have posted this in the Technical Architecture topic area...
My view is when dealing with fact tables that have degenerate dimension attributes I should create a 'Fact Dimension' in SSAS which contains the DD attributes only and it will share the same physical table as the fact data measures. Any atributes which are deemed as 'Regular Dimensions' (such as the ProductID in your example) should have it's own discrete dimension table which has the benefit of being able to implement slowy changing dimension functionality. This highlights a problem with 'Fact Dimensions' in that they can't implement SCD2 since they are directly bound to fact in which case you are better of treating them as 'Regular Dimension'
My view is when dealing with fact tables that have degenerate dimension attributes I should create a 'Fact Dimension' in SSAS which contains the DD attributes only and it will share the same physical table as the fact data measures. Any atributes which are deemed as 'Regular Dimensions' (such as the ProductID in your example) should have it's own discrete dimension table which has the benefit of being able to implement slowy changing dimension functionality. This highlights a problem with 'Fact Dimensions' in that they can't implement SCD2 since they are directly bound to fact in which case you are better of treating them as 'Regular Dimension'
Ham09- Posts : 8
Join date : 2011-07-26
Location : United Kingdom
Re: SSAS 'Fact Dimension' v 'Regular Dimension' your views
Fact dimension (aka. DD) in SSAS is natural key by itself, so the concept of SCD2 is not applicable to DD. Think of it this way, suppose you want to store it in a single attribute dimension table and have FK in the fact table, your dimension can be nothing but type1 dimension.
hang- Posts : 528
Join date : 2010-05-07
Location : Brisbane, Australia
Re: SSAS 'Fact Dimension' v 'Regular Dimension' your views
Thanks hang, I have given you my take on your other post in this topic area :-)
Ham09- Posts : 8
Join date : 2011-07-26
Location : United Kingdom
Similar topics
» Too many Time Dimension Role Views?
» Dimension relation Many-many and regular relation
» How to model 4 parallel views/hierarchies of a dimension?
» Derived fact tables, aggregation and views
» Montly Snapshot views with transactional Fact table w/SCD2
» Dimension relation Many-many and regular relation
» How to model 4 parallel views/hierarchies of a dimension?
» Derived fact tables, aggregation and views
» Montly Snapshot views with transactional Fact table w/SCD2
Page 1 of 1
Permissions in this forum:
You cannot reply to topics in this forum