Administering Analysis Services - Professional SQL Server Analysis Services 2005 with MDX - page 621

Analysis Services and Fail-over Clustering
Analysis Services 2005 provides fail-over clustering out of the box. By that we
mean SQL Server 2005 setup supports Analysis Services to be installed on a
windows clustered environment. You first need to have Microsoft Cluster Ser-
vices (MSCS) setup to form a cluster of two or more nodes. We recommend
you look at Windows product documentation to setup and configure MSCS.
You need to have a shared disk with sufficient disk space that can hold the
Analysis Services data files. After setting up a cluster; you can then use Mi-
crosoft SQL Server 2005 setup to install Analysis Services with fail-over cap-
ability. In the SQL Server 2005 setup you need to provide a virtual server
name with an IP (internet protocol) address and the shared data folder for the
Analysis Services data. The virtual server name will be the name of the Ana-
lysis Services server. SQL Server 2005 setup installs Analysis Services 2005
binaries on all the nodes of the cluster and the data folder is setup on the
shared disk with appropriate permissions.
Figure 12-33 shows the overall architecture of the Analysis Services 2005 fail-
over cluster setup. Users of Analysis Services will only be aware of the virtual
server name. At a give time only one of the physical machines will service
users requests and has access to the shared disk containing Analysis Services
data. When one of the nodes in the cluster fails due to network or power prob-
lems then MSCS makes the second node as the primary and provides control
to the shared disk. MSCS identifies failure in the primary node through the
heart beat which is a typically a connection through a second network card
between the machines involved in the cluster. The second node gains control
to the Analysis Services data all future user requests are directed to the
second node. Existing users originally connected to the first node would have
to re-establish their connections. In order to make sure you are isolated from
disk problems it is recommended that the shared disk is a RAID (Redundant
Array of Independent (or Inexpensive) Disks to provide fault tolerance. Due to
fail-over clustering support you have higher availability of Analysis Services to
your users whenever there are hardware problems.