Weighting

• Another problem with range partitioning is that a join attribute may appear a different number of times in different partitions.
• Eg. A relation R = {1,2,3,4,4,4,4,4,4,4,4,5} to be partitioned over 3 processors. An even partitioning vector corresponds to [4,4]. Hence to balance the load evenly among the 3 processors, 1/8 of the tuples containing 4 as the join attribute (along with join attribute values 1,2,3) must be directed to P0, 1/2 of the tuples containing 4 must be directed to P1, and 3/8 of the tuples containing 4 (along with join attribute value 5) must be directed to P2.
• This is known as weighted range partitioning.