Planning broker capacity

This scenario describes how you can use the WebSphere® Message Broker Monitoring Agent to plan the capacity for your brokers.

To determine the most recent message that was sent between two applications, do the following steps so that the WebSphere Message Broker Monitoring Agent can gather the necessary statistics:

  1. Place one CandleMonitor node immediately following the MQInput node so that the entry of the message into the message flow is recorded.
  2. Set the type attribute of the CandleMonitor node to input.
  3. Place one CandleMonitor node immediately in front of the MQOutput node so that the exit of the message from the message flow is recorded.
  4. Set the type attribute of the CandleMonitor node to output.
  5. Start the WebSphere Message Broker Monitoring Agent again.

After you have positioned the CandleMonitor nodes and the WebSphere Message Broker Monitoring Agent is running, do the following steps:

  1. Open statistics workspaces, such as Monitor Node Message Flow Statistics/Message Flow Statistics for individual message flows, Monitor Node Execution Group Statistics/Execution Group Statistics for summarization over an entire execution group, or Monitor Node Broker Statistics/Broker Statistics for statistics summarizing the operation of an entire broker.
  2. Examine the statistics workspaces and their historical workspaces (which you can link to from the statistics workspaces).
  3. If you find that for a particular message flow the Current Message Input Rate value is higher than expected, or that the Current Average Queue Time value is unacceptably long, add more instances to the flow in the Message Brokers Toolkit.
  4. If you find that for a particular execution group the Current Message Input Rate value is higher than expected, or that the Current Average Queue Time value is unacceptably long, add a new execution group and assign instances as appropriate.
  5. If you find that for a particular broker the Current Message Input Rate value is higher than expected, or that the Current Average Queue Time value is unacceptably long, consider adding a new broker with a configuration that can offload some processing from the current broker.