As a consequence, there is no need for distributed query
processing but instead each node of the cluster is capable of evaluating any OLAP
query stand-alone and in parallel to other nodes. The decision on which node to
actually use for query evaluation is called query routing.
The objectives of query routing are to balance the load of the cluster nodes and to
reduce query response times. Effective query routing requires some knowledge about
the current state of the cluster. In particular, this comprises the available nodes and
the number of currently active transactions at each node. In the following, we refer
to the latter as the load of a node, also known as current degree of multiprogramming
(Weikum & Vossen, 2001).
OLAP with a Database Cluster 2
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
A middleware-based approach has the advantage that global knowledge about the
current system state can also be easily maintained if all clients issue their queries
to the coordination middleware of the database cluster. In such a middleware-based
architecture, query routing is actually twofold. Clients place their queries into the
input queue of the cluster coordinator. In our terminology now, routing of a (single)
query is the decision on which node of the cluster shall execute a query.
Pages:
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451