Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
the data through UNION ALL views and distributed SQL (Microsoft calls this technology
Distributed Partitioned Views DPVs) (Otey & Otey, 2005). The DPV is constructed differently
at each node. It must explicitly consider which partitions are local and which are remote.
Figure 2 shows the overall structure of the shared nothing database clusters architecture.
Shared.Disk.vs..Shared-Nothing.Storage.
This section will try to demonstrate the main variations in implementation aspect between
the shared and shared nothing storage approaches. The discussion will focus on application
development and database high availability concerns.
??? Application.development:.Generally, database clusters appear just like a regular database,
which means there are no additional constraints on the application developer. An
application, even a complex OLTP for a Symmetric Multiprocessing (SMP) platform,
runs with no modification on a shared-disk cluster. The single database image, with
no data partitioning, carries over from the SMP to the cluster.
Figure 2. Shared nothing database clustering architecture
Database
Partition
Database
Partition
Database
Partition
Database
Partition
Database
Partition
Database
Partition
Rada deh & Al-Ameed
Copyright ?© 2007, Idea Group Inc.
Pages:
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95