Type of Dimensions
Conformed Dimension:
When a dimension can be used with multiple Facts, more precisely with multiple facts from multiple Data Marts. Conformed dimension provide the ability to perform the analytics across multiple subject ares of the Enterprise.
Most common example is Date Dimension.
Others are Product Dim, Customer Dim.
Junk Dimension:
Its a way to store flags and indicators that are associated to transnational data. By creating Junk dimension these flags and indicators can be removed from the corresponding Fact table and placed in Junk dimension.
The transaction table (in OLTP) may have several yes/no, true/false, indicator or comments fields or columns with low carnality values.
If we leave these in Fact table then the row length would be unnecessary large (and so the I/O).
While creating a respective Fact table in Data Warehouse, one way is to those columns in the Fact table as it is, other way is use a junk dimension that holds all the unique combinations of those indicator fields into a single dimension and assigns a unique key. This key is what is stored in the fact table. So you will have only one additional dimension table and will reduce the number of fields in the fact table.
Lets say there is Order Transaction Table that contains flags like Created (yes/no), Shipped(Yes/no), Returned ( Yes/No), Cancelled (Yes/No). These flags are related to status of the Order (or Order Line). There would be Order Key in the Fact Table that is linked to Order Dimension. There is no need to store Flags in the Fact table. In this case there is no need for Junk Dimension.
But at times there could be flags that are not associated to any dimension, and if there are plenty then keeping them in a Junk dimension would be more appropriate rather than storing them in the Fact table.
(Let me find out some real life example of such flags and indicators and I will update here)
