"Edward Atkinson" wrote:
> I was just wondering how you calculated that each dimension table in scenario C for the example you ran through in lecture 4 would have 60,000 possible rows?
To answer my own question in the hope it might help someone else. The 60,000 unique values for each dimension limitation is a product of the fact that a 2 byte key is used to index rows in the dimension tables. This means that 2^16 - 1 rows can be stored, 65536 rows, therefore giving us our 60,000 unique value limit.
> Wouldn't it be possible for a dimension to have 100 million unique values if there are 100 million facts in the fact table? How can you be certain there would be only 60,000 possible values for each dimension, or is that simply an assumption?
In the context of scenario A the assumption that there are only 60,000 unique values for each dimension also still makes sense as the likelihood of two rows in a dimension table of 100 million rows having the same values for 100 attributes with 60,000 possible levels is practically 0.
Lecture 4: 60,000 possible dimension table rows (all 3)