For segmenting on big tables it is common practice and recommended to use primary key as hashing key.
Now for two big tables A, B for which a join query on condition A.PK=B.FK_A is frequently performed (where B.FK_A is B’s FK to A), if A and B are both segmented on their own PKs, since A.PK and B.PK are different, there is big chance that the matched records of the given query are not in the same nodes.
If instead, let A be segmented on A.PK, and B be segmented on B.FK_A (instead of on B.PK), then all matched records for the given query in question will be in the same respective nodes; and so the performance of the query may be better. In general, using FK as segment key is sometimes helpful, e.g., in the above scenario.
(Here we may think of A, B as for transaction and transaction detail respectively, for example)
Is above thinking in right track, and comments?
