Description

When your application must support large numbers of users, high data volumes or is required to be fault tolerant, you may need to run several AltioLive Presentation Servers in a cluster. A cluster may be configured in several ways, depending on your needs. See Example Architectures for some sample cluster configurations.

Clustering for Performance

The simplest way to handle large numbers of users or high volumes of data is to direct different users to different Presentation Servers which are working together. Depending on the load profile, Presentation Servers may have peer relationships or Master/Slave relationships with each other. A single Presentation Server cluster may contain Presentation Servers with both types of relationship. For example, if the back-end is supplying data that needs to go through a complicated XSL transformation, the Master may handle the transformation, off loading the users to Slave Presentation Servers, thus spreading the back-end and front-end loads.

When configuring a cluster, you must list all the servers in each server's configuration. This is best done with the Presentation Server Administration Tool, but can also be configured by editing altioserver.xml on each machine directly. When clients access the Presentation Server through a redirector, sometimes the Presentation Servers will not be able to identify which of the configured servers is the local one, and will fail to start properly. If this happens, it is necessary to edit the altioserver.xml file on each server by hand, adding a THIS attribute to the SERVERS element identifying which SERVER is the local one.

When setting up datapools, the slave settings must also be configured when you have Master/Slave relationships in your cluster.

Clustering for Fault Tolerance

When fault tolerance is needed, in addition to clustering at the Presentation Server level, you will also need to cluster the Application Servers that the Presentation Server is running on, so that users' session information is shared within the cluster. In addition, some form of network redirector will be required to transparently reroute requests when a server goes down. In this configuration, it is almost always necessary to add a THIS attribute to the SERVERS element as described above.