Question

We have discussed parallel query processing on a parallel database system with n nodes, with a focus on join processing. Suppose that there are two relations, r and s, where the schema of r and s are R(A, B) and S(A, B), respectively. Consider two ways to partition the two relations on all n. (a) The two relations are partitioned on all n nodes using round-robin parti- tioning. (b) The two relations are partitioned on all n nodes using either range- partitioning or hash-partitioning. Consider if you can parallelize the following queries when relations are parti- tioned by (a) and (b), respectively. Hint: You need to give the partitioning attribute name(s) with which the relations are partitioned. You need to con- sider how to partition relation r and/or s on all n nodes where $r_i$ and $s_i$ are the partition at the $i$-th node, consider how to process the operation on the $i$-th node locally, and consider if the union of all the local results at all nodes is the same as the final answer. 1. the difference operation such as $r - s$, (select * from r) except (select * from s) 2. aggregation by the count operation select count(B) from r group by A) 3. aggregation by the count distinct operation select count(distinct B) from r group by A)

          We have discussed parallel query processing on a parallel database system with
n nodes, with a focus on join processing. Suppose that there are two relations,
r and s, where the schema of r and s are R(A, B) and S(A, B), respectively.
Consider two ways to partition the two relations on all n.
(a) The two relations are partitioned on all n nodes using round-robin parti-
tioning.
(b) The two relations are partitioned on all n nodes using either range-
partitioning or hash-partitioning.
Consider if you can parallelize the following queries when relations are parti-
tioned by (a) and (b), respectively. Hint: You need to give the partitioning
attribute name(s) with which the relations are partitioned. You need to con-
sider how to partition relation r and/or s on all n nodes where $r_i$ and $s_i$ are the
partition at the $i$-th node, consider how to process the operation on the $i$-th
node locally, and consider if the union of all the local results at all nodes is the
same as the final answer.
1. the difference operation such as $r - s$,
(select * from r) except (select * from s)
2. aggregation by the count operation
select count(B) from r group by A)
3. aggregation by the count distinct operation
select count(distinct B) from r group by A)

We have discussed parallel query processing on a parallel database system with
n nodes, with a focus on join processing. Suppose that there are two relations,
r and s, where the schema of r and s are R(A, B) and S(A, B), respectively.
Consider two ways to partition the two relations on all n.
(a) The two relations are partitioned on all n nodes using round-robin parti-
tioning.
(b) The two relations are partitioned on all n nodes using either range-
partitioning or hash-partitioning.
Consider if you can parallelize the following queries when relations are parti-
tioned by (a) and (b), respectively. Hint: You need to give the partitioning
attribute name(s) with which the relations are partitioned. You need to con-
sider how to partition relation r and/or s on all n nodes where ri and si are the
partition at the i-th node, consider how to process the operation on the i-th
node locally, and consider if the union of all the local results at all nodes is the
same as the final answer.
1. the difference operation such as r - s,
(select * from r) except (select * from s)
2. aggregation by the count operation
select count(B) from r group by A)
3. aggregation by the count distinct operation
select count(distinct B) from r group by A)

Added by Inmaculada K.

Question

Please give Ace some feedback