simple and dumb question on index

[harry sheng]

Say I have a fact table with 3 columns as primary key, and 1 column for value.
One of the primary key column is datetime type, all other columns are BIGINT type.

One of the primary keys has 2k unique values, another has 16 distinct values, and every datetime value repeats 16k to 32k times in the table.

The total number of records may grow up to more than 100M.

I need to support very selective query and not-so-selective (say return 1/100th to 1/10th of the whole table) query.

Let's put the disk requirement and update time performance aside, will there be any performance benefit for query if I add an index which include all the columns, the key coloumns and the value column? I mean, I get everything from the index, I do not need to go to the table, right? And the index size will be much smaller than the table, that's the argument point here.

I know this is an odd case and an odd question. Just curious.

Thank you very much.