Date Dimension or DateTime Field

Hi,

I'm working on a project where dimensions have dates in them. An example is for a client where we have date of birth, joining date, date of death, etc. In fact i have about 6 different date columns.

Now, I'm trying to find out what the best practice is for this, do I set them as datetime fields and only add the actual date, or do I link them to the calendar dimension?

I've searched the internet, read my Kimball texts and haven't been able to come up with an answer yet.

Using datetime keeps a very clean star schema, however, at a greater cost of storage in comparison to integers.

Linking them to the calendar would make for faster searching on these fields and to be able to query on the different calendar attributes, months, years, etc... It would also use less storage. But, it creates a snowflake schema presenting a far more complicated model.

So I'm curious to hear how other people are dealing with this situation and to try and find out what the best practice actually is.

I use snowflakes if there is any chance that reporting will want to leverage the time dimension or if I need to track time for the date. You could always push the dates down to the fact table if they are applicable.

So basically, if there is a possibility that the business might want to query on the date, ie. break reports down to what year clients joined, use the date/time dimension, otherwise just use the datetime field.