Comentários (2)

1 asavory comentou às Link permanente

For Data Warehouse environments I've also seen a best practice recommendation that states that fact tables should be distributed using the foreign key of the largest dimension. The logic here is that the largest dimension will not otherwise be collocated with the fact data and is likely to be the most expensive join. This makes the following assumptions:

 
1. The dimension is too big to be replicated.
2. The join to the dimension is performed (i.e. fact queries reference the dimension) as much as joins to other dimensions.
3. There are no other big tables (e.g. associated fact tables with a common primary key) that may be joined more frequently.

2 sboivin comentou às Link permanente

Thank you for your comment Adrian. The is a useful recommendation, and I'll forward it to the paper's authors to make sure they are aware of it.

Incluir um Comentário Incluir um Comentário