Configuring an Engine

Governor The governor does not do much processing and does not require a high performance server. It is recommended that you not assign it to the same server as either the primary or backup engine unit, because it would then not be able to distinguish between failure of the primary engine unit and a break in the network link between the engine units.

Database Services The number and placement of database services depends on load conditions. You must have enough database services running to prevent database access requests from backing up in the database service router (see “Viewing Performance Indicators” on page 222). In general, assign high priority database services to a very capable server node, probably the same as the one on which the database manager resides. If your database system provides network access to the database, such as ORACLE’s SQL*Net, then lower priority database services can be assigned to under-utilized servers, where resources are available to help carry heavy loads on the engine. (If your database system does not provide network access to the database, then your database services must reside on the same node as your database manager.)

A typical configuration is illustrated in Figure 4-2 on page 88.

In this illustration the database services on Server4 are used in a round-robin fashion unless the load on the engine is too heavy for them to carry it. In that case the database service on Server1 is used. The database service on Server3 will be used to handle peak loads, but will get less work than the service on Server1.

You did not have to decide at iIS installation time exactly which engine components would be assigned to each server node because all component partitions are installed on each server node that supports an engine. Instead, you must select which components to start on which nodes when you configure each engine.

For example, you can designate a default set of database services that are started at engine startup time. Subsequently, however, you can start up additional database services on designated nodes, as needed, to accommodate increasing load on the engine.

The location of the primary and backup engine units is best determined in advance, before startup time, and should not change because the shared library files corresponding to process definitions registered with the engine are needed on all the nodes on which the engine units are located. For this reason, moving the engine units from one node to another can take a long time.

Your default engine component partitioning scheme is stored in an engine configuration file, allowing you to design your basic engine component partitioning scheme in advance and consistently start up your engine the same way every time.

Chapter 4 Managing Engines 87

Page 87
Image 87
Sun Microsystems 3 manual Configuring an Engine