IBM E02HMLL-I Implementation Guide - Page 292

Concurrent event-triggered flow processing in connector controllers Connector controllers can be configured to process multiple event-triggered flows concurrently. Event-flow processing performance improves when a connector is configured to process triggered event flows concurrently. This is because multiple business objects can be transformed in mapping at the same time. To configure concurrent event-triggered flow processing for connector controllers, you set the ConcurrentEventTriggeredFlows property to the maximum number of flows you want processed at the same time. For more information on this property, see "ConcurrentEventTriggeredFlows" on page 137. For more information on using Connector Configurator to set connector properties, see in Chapter 7, "Configuring connectors," on page 121 general. Tip: If the ConcurrentEventTriggeredFlows property is configured to a value greater than 1, the connector controller maintains the same order of events that it received from the application. Concurrent processing of event-triggered flows in connectors requires additional system resources. To maximize performance, ensure that the system resources that handle concurrent events are not idle. For example, do not set the ConcurrentEventTriggeredFlows property to a value of 10 if the number of events arriving at InterChange Server Express for the connector never reaches 10. Use the server statistics window to determine the number of events that are in connector's queue by monitoring the MQ queue depth for the connector. Monitoring this statistic can help you set the value for the ConcurrentEventTriggeredFlows property. If a connector is being used only as a destination, you can improve performance by setting the value of ConcurrentEventTriggeredFlows to zero. But do not set this value to zero if the connector is being used in bidirectional exchanges with a collaboration. Distributing connector agents You can improve the performance of the business integration system by distributing the connector agents. The reasons connector agent distribution improves performance include the following: v Each InterChange Server Express and connector agent process runs in its own Java Virtual Machine (JVM). A computer can only have so many JVM instances running on it before performance begins to degrade. The threshold is dependent upon the computer specifications, but distributing a connector agent decreases the load on the machine hosting InterChange Server Express by one JVM. v The performance of an individual connector generally improves the closer it is to the application with which it communicates, with respect to the network topology. Ideally the connector agent should be installed on the same computer that hosts the application server itself, though this might not be possible depending on the environment. If the application host computer and the InterChange Server Express host computer are on different subnets, however, the connector agent will typically still perform better if installed on another computer in the same subnet as the one in which the application host computer exists. For instructions about setting up distributed connector agents, see the System Administration Guide. 280 IBM WebSphere Business Integration Server Express and Express Plus: System Implementation Guide