IBM E02HMLL-I Implementation Guide - Page 291

Chapter 14. Performance tuning This chapter describes how to implement various techniques and configurations to improve system performance. It contains the following sections: v "Implementing concurrent processing of event-triggered flows" v "Caching static relationships" on page 281 v "Using database connection pools" on page 281 v "Using the memory checker thread" on page 281 Implementing concurrent processing of event-triggered flows You can configure collaboration objects and connector controllers to process multiple event-triggered flows at the same time. This can significantly improve the performance for event-triggered interfaces. Concurrent event-triggered flow processing in Collaborations Collaborations can be configured to process multiple event-triggered flows concurrently. System throughput and response time to event processing improve when concurrent event processing is properly used (see tip below). By default, a collaboration processes one event-triggered flow at a time. When a collaboration is processing concurrent event-triggered flows, the collaboration identifies dependencies among those flows and processes them in the order that they were sent from the connector controller. Concurrent processing of event-triggered flows is performed on flows that do not have data conflicts, while flows with data conflicts are processed in the order received. To configure a collaboration to handle multiple event-triggered flows, see "Maximum number of concurrent events" on page 167. Tip: Concurrent processing of event-triggered flows in collaborations requires additional system resources. To maximize performance, ensure that system resources used to handle concurrent events are not idle. For example, do not set the value for the Maximum number of concurrent events to 10 if the number of events in the collaboration queue never reaches 10. If a collaboration's inbound ports are bound only to receive external calls through the Server Access Interface, and are not bound to any connectors, you can improve performance by setting the value of Maximum number of concurrent events to zero. But do not set this value to zero if the collaboration is being used for bidirectional exchanges with connectors. Concurrent event-triggered flow processing in collaboration-object groups Each collaboration in a collaboration-object group can be configured independently for processing a number of concurrent event-triggered flows. It is recommended that you set the same value for the number of concurrent event-triggered flows for all of the collaborations in a group, so that a collaboration with a low concurrency rate does not become a bottleneck. © Copyright IBM Corp. 2001, 2004 279