Thanks for the article.
Just to emphasize the Hash table section some more. There is only 60 distributions that your Hashed key will be stored in. That means you should at least have 60 unique values or more with the column of data you choose to hash on in a deterministic fashion.
As mentioned in the article, if you have a lot of default values or NULLs in that particular field, then you may want to consider another column or blend those defaults in some manner that can evenly be distributed across those 60 distributions. In the event you ignore the default value, then you can end up with one database of the 60 with all the values for that particular default value. Thus, if you ever query it, then you end up with one database doing all the work within the compute node versus distributing the compute across the other nodes.'