FileNet P8 Content Search Engine, Version 5.2.1            Operating systems:  AIX, Linux, Linux on System z, Solaris, Windows

Choosing a load balancing method for IBM Content Search Services servers

To optimize indexing and search performance, you need to decide on a method to balance the load among the IBM® Content Search Services servers.

About this task

By default Content Platform Engine uses a built-in load-balancing algorithm to assign IBM Content Search Services servers to indexes according to the indexing workload of the servers. The assignments are based on the number of index servers and the resources that are available to each server. If you want to override this built-in algorithm, you can use Administration Console for Content Platform Engine to create affinity groups and manually dedicate IBM Content Search Services servers to specific index areas.

An affinity group is a group of one or more servers that are dedicated to one or more index areas. A server that is a member of an affinity group can serve only index areas that are assigned to that affinity group and that belong to the same site as the server. A server that is not a member of an affinity group can serve only index areas that do not belong to an affinity group and belong to the same site as the server.

With an affinity group, the administrator can limit the load balancing for an index area to the servers that are members of the group. These servers do the indexing and searching of full-text indexes. All servers in the group must have equal access to the root directory of the index area. The affinity group should include servers that can do indexing and searching.

The affinity group improves performance because you can index your data on a disk that is local to IBM Content Search Services. The downside is that Content Platform Engine cannot provide failover. If the local disk that hosts the index area fails, all indexing and search requests to that index area fail.

To avoid the possibility of a single point of failure for an affinity group, do not store full-text index data on local (non-shared) disks. Instead, store your index data on shared disks with data redundancy, as described in IBM Content Search Services distributed installation scenario.

If you must use local disks, be sure to implement data redundancy by using a high availability strategy for failover of the IBM Content Search Services server and the disks, provided by Veritas, Microsoft Cluster Server, or IBM PowerHA®.



Last updated: March 2016
p8ppi255.htm

© Copyright IBM Corporation 2013, 2016.