I agree with Sergiy on the indexes on datetime data types.  These are just represented by integers anyway, aren't they?  Is there a large additional overhead because of using datetime over, say, integer?

I have a table that logs requests from users and will be frequently queried on the request date (either one day or a date range).  I have used a non-clustered index on an IDENTITY PK and a clustered index on the request date (smalldatetime).  The likelihood of two requests being made during the same minute is quite small so the request date column is almost unique.  Also the request date inserted will always be greater than or equal to the greatest request date.

The table will start at 0 and will only increase by about 2-3000 rows each year so I doubt the index options would make much difference. 

Does anyone have some more info on the use of datetime/smalldatetime as indexes, clustered or non-clustered?