An overview and framework for workload allocation in distributed database systems
is presented in Rahm (1992). Approaches to affinity-based transaction routing
typically concentrate on OLTP-like transactions. In particular, they only consider
the data accessed by the queries to decide whether there is affinity between two
given queries. The nature of the access is not taken into account, that is, ???scan??? vs.
???random access.??? A strong disadvantage is that such approaches rely on precomputed
data from low-level trace information of previous runs of the workload. This
neglects the component-oriented nature of a database cluster.
OLAP with a Database Cluster 24
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
Cache.Approximation.Query.Routing
The objective of query routing is to reduce query response time. The execution
times of queries??”especially OLAP queries??”are dominated by I/O costs. While
caching plays a major role for performance of query evaluation, all previous mentioned
routing algorithms do not take caching effects into consideration at all. In
contrast, cache approximation routing estimates the benefit of a cache state for a
query. The idea is that the execution time of a query is minimal at the cluster node
whose cache contains the largest subset of data that will be accessed by the query
(R?¶hm, B?¶hm, & Schek, 2001).
Pages:
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455