.Results.merging:
Select sum(suma), sum(counta),
sum(suma) / sum(counta), max(maxa), min(mina)
(sum(ssuma)-sum(suma)2)/sum(counta), group_
attributes
From UNION_ALL(PRqueryX), dimensions
(join)
Group by group_attributes;
Efficient and Robust Node-Partitioned Data Warehouses 2
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
tive. The expression 2 1 R A R ?????? is processed as ( 21 11 R A R ?????? ) U ??¦ U ( n R A n R 2 1 ??????
), each part of this expression in a different node, because as the two relations are
partitioned by the equi-join key, the join between two fragments in different nodes
is an empty set (e.g., = 22 11 R A R ?????? ). Otherwise, at least one of the relations must
be moved. If only one of the relations or neither is partitioned on the join key, we
can dynamically repartition on the same join key and proceed with the parallel equijoin
??”this is the ???Repartitioned.Join??? (RpartJ). The repartitioning is accounted as
an extra overhead, which increases total work and response time and is dependent
on data buffering and communication-related overheads. On the other hand, if one
of the relations is replicated by placement, the join can proceed independently at all
nodes regardless of the partitioning key for the other relation.
Pages:
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408