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

Installing and configuring IBM Content Search Services

You can install and configure IBM® Content Search Services for a single server configuration or a multiple server configuration on Windows, AIX®, Linux, and Solaris operating systems.

Content Platform Engine preprocesses documents and sends them to the IBM Content Search Services index server for indexing. Depending on the size and configuration of your system, you might want to have multiple server instances of IBM Content Search Services installed on your system.

IBM Content Search Services server performance degrades when indexing and searching run in parallel (dual mode) under moderate to high content-based retrieval (CBR) workloads. You can achieve optimal performance under moderate to high CBR workloads if you dedicate each IBM Content Search Services server to either one or the other operation (searching or indexing) and run each dedicated server on a separate physical machine. That is, each server should have its own dedicated CPU and memory resources. For detailed information on optimizing searches and indexing, see http://www.ibm.com/support/docview.wss?uid=swg21597423

In a single server configuration, a single instance of the IBM Content Search Services server performs both the indexing and searching tasks for the Content Platform Engine.

In a multiple server configuration, you assign different roles to your IBM Content Search Services server instances to distribute the load for both indexing and searching tasks. You can run multiple server instances of IBM Content Search Services on the same computer for vertical scaling or you can run them on different computers for horizontal scaling. Each server instance of IBM Content Search Services can be designated to do only indexing tasks, searching tasks, or both. You designate which tasks each IBM Content Search Services server does by assigning it a specific server mode:
Index
Content Platform Engine sends only indexing tasks to servers in this mode. You can have multiple, dedicated index servers to distribute the indexing load for high volume scenarios. Multiple index servers also allow failover capability for the Content Platform Engine. If one indexing server goes offline, the Content Engine automatically distributes the indexing load to the remaining indexing servers.
Search
Content Platform Engine sends only searching tasks to servers in this mode. You can have multiple, dedicated search servers to distribute the searching load when you have many indexes to search. Multiple search servers also allow failover capability for the Content Platform Engine. If one search server goes offline, the Content Engine automatically selects an alternate search server to perform its tasks.
IndexAndSearch
Content Engine sends both indexing and searching tasks to servers in this mode. The IndexAndSearch server mode is typically used in a single server configuration.
Tip: When considering how to balance available CSS servers, it is important to take into consideration the expected indexing and search scenario. The following table presents possible scenarios and the suggested server mode assignments. Note that even a heavy search load is generally lighter compared to the indexing/ingestion load.
Table 1. Choosing the server mode
Expected scenario Suggested server mode
A heavy indexing load (for example, an indexing rate of two million documents per day) Configure more CSS servers in Index mode.
A heavy search load Configure one or more servers in Search mode. If you are limited to two or three servers, configure one server in Search mode and the other servers in Index mode or IndexAndSearch mode.
A heavy indexing load and a light search load, and you are limited to two or three servers Configure all CSS servers in IndexAndSearch mode.

Set up the site to contain the object store and the IBM Content Search Services servers. All of the documents and other objects are indexed in the same site as the object store.



Last updated: March 2016
p8pin273.htm

© Copyright IBM Corporation 2013, 2016.