OLAP
The overall goal is to achieve high scalability and performance for OLAP queries
and to allow OLAP clients to access up-to-date data if they ask for it. In order to
achieve this, we drop the strict distinction between operational databases and data
warehouses. Instead of separating OLTP and OLAP workloads in space, we propose
to deploy a middleware-based database cluster for both workloads simultaneously
as illustrated in Figure 1. The ratio behind this unified architecture is to combine
the scalability and performance of parallel database clusters with quality of service
guarantees possible through a coordination middleware.
Database clusters are comprised of a cluster of commodity computers as hardware
infrastructure and off-the-shelf database management systems as a transactional
storage layer. Such a database cluster is an attractive platform for both OLTP and
OLAP with regard to performance, scalability, fault tolerance, and cost/performance
ratio. Several transactions can run in parallel on several nodes. When the workload
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.
or the amount of data increases, one can ???scale out??? the cluster by simply adding
new components (Gray, 1999). Finally, fault tolerance can be achieved using replication
of either hardware or data.
Pages:
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439