After you determine the hardware configuration, determine how to deploy Sybase Search across multiple servers. Ideally, there should be no more than one container per server. Multiple containers residing on a server must access the same disk drive, which can slow down performance. The Java 2 Platform Enterprise Edition (J2EE) server hosting the Sybase Search Web application should also reside on its own server. Distributing Sybase Search across multiple servers helps maximize the resources available to each container and helps prevent I/O bottlenecks.
Installing one container per server also reduces unnecessary network traffic among containers. If two containers are installed on one server, their network traffic can be eliminated by combining their internal modules together into a single container. There is no advantage to running more than one container on a single server; it is not recommended.
For a small installation on one server only, Sybase recommends that only one container is used, with the modules often shared across multiple containers located together in a single container.
Each container runs within a Java Virtual Machine (JVM)
tied to a single CPU. It is possible to run multiple containers
on a server with multiple CPUs, with each container’s JVM
attached to a different CPU.
A clear understanding of performance expectations and how many documents you plan to index help determine how many servers are needed in your environment. Table 2-2 shows an example.
Setup |
Sybase Search feature |
Number of servers |
|---|---|---|
Small: Less than 500,000 documents are planned to be indexed. |
Single server installation |
One server |
Medium: 500,000 to 1.5 million documents are planned to be indexed. |
Hub container Satellite container Web application |
One server One server One server |
Large: More than 1.5 million documents are planned to be indexed. |
Hub container Satellite container Web application |
One server Five servers One server |
