Hello,
that's right, updating the PI column(s) for a row means that the row will be re-distributed to another AMP (unless the target AMP is the same). So, it is good to choose PI column(s) that is not updated frequently. Some updates are fine, but not many rows very frequently.
Multiple columns in PI - yes, it means that the combination of columns will be used to decide the AMP number.
You can use the following SQL functions to understand what AMP number is assigned to any value or combination of values:
select hashamp(hashbucket(hashrow(<column1_value>)))
select hashamp(hashbucket(hashrow(<column1_value>,<column2_value>)))
Hello,
that's right, updating the PI column(s) for a row means that the row will be re-distributed to another AMP (unless the target AMP is the same). So, it is good to choose PI column(s) that is not updated frequently. Some updates are fine, but not many rows very frequently.
Multiple columns in PI - yes, it means that the combination of columns will be used to decide the AMP number.
You can use the following SQL functions to understand what AMP number is assigned to any value or combination of values:
select hashamp(hashbucket(hashrow(<column1_value>)))
select hashamp(hashbucket(hashrow(<column1_value>,<column2_value>)))