If a
node in the shared disk cluster fails, the system dynamically redistributes the workload
among all the surviving cluster nodes. This ensures uninterrupted service and balanced
cluster-wide resource utilization.
On the other hand, data in federated databases is divided across databases, and each database
is owned by a different node. The only way to access data owned by a node is to request
the data from the node, and have the node service the request. Thus, when the node fails,
the data that it owns becomes unavailable, and the entire system becomes unavailable as
well. Also, any in-flight distributed transactions controlled by that node might have locked
data on other nodes. Therefore, recovering from node failures requires additional work in
resolving these in-flight transactions.
As instances of a recently-provided database clustering commercial solutions, Oracle Real
Application Clusters have started to address cluster architectures using shared storage
Database H gh Ava lab l ty: An Extended Survey
Copyright ?© 2007, Idea Group Inc. Copying or distributing in print or electronic forms without written permission
of Idea Group Inc. is prohibited.
systems such as SAN (Storage Area Network).
Pages:
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98