Ask the Experts Replay: WebSphere Message Broker - Message Flow Design Strategies And Usual Pitfalls



This WebSphere Support Technical Exchange Ask the Experts session addresses general WebSphere Message Broker V7 message flow design strategies and usual pitfalls that one can get into in the areas of design, implementation, configuration and testing.


Ask the Experts sessions are conducted in a question and answer format. The table below provides a time index (minutes:seconds) to the recording and describes the questions or topics discussed. You can fast forward to any question using the time index, a table containing the panel of experts is also included.

To play or download the audio of this Ask the Experts session, see the Audio Section of this document.

See the Related Information Section of this document for a list of documents referenced during the presentation.

Ask the Experts session
20 October 2011 - 11:00 a.m. - 12:00 p.m. EDT

Time Questions asked
00:00 Silence
00:00 General introduction
03:39 Technical introduction
04:58 What is the best way to find out why my message flow is slow or why it is taking lots of CPU?
18:30 Which has better performance: multiple inputs in one flow or different flows?
19:04 Can you comment on the use of user defined notes and performance?
19:04 Can we have one message flow with MQ Input node, JMS input node, SOAP input node and HTTP input node - followed by a compute node (i.e. a message flow which can take input on different protocols)? If yes, are there any limitations we should be aware of?
20:03 Is there a sandbox environment we can play in to learn more about WebSphere Message Broker? Are there any good "how to" websites for WebSphere Message Broker?
23:57 What are the possible pitfalls for using message flow monitoring to audit and monitor my message flows?
29:57 What is the best monitoring tool to monitor the performance of message flows without slowing them down?
31:28 Is there any way we can find out about the status of a tracing on the Broker?
32:35 How can WebSphere Message Broker validate a message against a schema and still have the ability to control the message direction so that you can send a good or bad response message?
35:43 How can you capture an invalid response from a SOAP request node?
36:31 What are the limitations of the Java compute node? Can I do everything I can in Java in a Java compute node?
40:17 What approaches are available to cache data from a database across flows in an execution group?
43:26 What is the best way to handle HTTP time out message when the flow exposes the web service?
44:09 What is the best option to outbound file transfers via FTP to a message flow without using a Java compute node?
44:33 Is there any built in functionality to retry a SOAP input node?
45:08 How do you refresh those cache nodes?
45:36 Are there any guidelines on handling large messages in Broker?
51:50 Is there any time based node with a terminal?
52:29 On flow with a commit count greater than one, how do I know which message in the bunch caused the error?
53:03 Is it possible to schedule a flow to trigger every day at the same time?
53:31 How do you suggest handling a requirement to implement pub/sub using files instead of messages?
54:18 Can we stop a message flow from processing a file if the file is larger than the expected size?
54:36 How do I take advantage of the new capability in V7 to setup unique environments for each z/OS broker execution group?
58:15 What are the limitations that one needs to understand in implementing message flows in Multi-instance Message Brokers? Are there any additional advantages with multiinstances on top of HACMP?
63:32 How do I get the retry count if I opted for the retry option?
63:59 Is there a best practice to determine how many execution group should be created for a distributed environment?
65:31 When I do PSAUS to my data flow engine, a particular EG shows memory utilization as 342MB when the JVM max heap size of that EG is only 256MB, what does the memory given on the PSAUS command account for?
66:57 Going back to batch mode for file nodes, even though the processing stops at the first error in the file the partial files created in the MQSI transit folder are not deleted - is there a way to get rid of them automatically?
68:10 Why do we throw an unit exception in a Java compute node?
68:58 Each time we configure a user data source entry on the ODBC, we use MQSI setDB prompt to set user and password, we need to cycle the Broker to take that into effect - can this be avoided?
69:24 What is the easiest way to increase/decrease the message flow instances at run time based on the transaction volume?
70:43 Can the validate node reference an external schema?
72:04 In V7+ can we change the number of instances from the MBExplorer without having to redeploy or restart the flow?
72:20 Does changing the CSD or the Fixpack level need a flow or execution group restart?
72:32 Are the additional instances a redeploy or a restart?
73:00 Does the change to the Java code require a full Broker restart?
73:42 Closing remarks
74:36 End of Call

Panel of Experts:
Tim Dunn WebSphere Message Broker Performance Architect
David Coles WebSphere Message Broker L3 Service Specialist
David Crighton WebSphere Message Broker L3 Service Specialist
Randy Miller WebSphere Message Broker L2 Specialist
S. Rao Nanduri WebSphere Message Broker L2 Specialist



To replay the recording of this 73 minutes conference call, click on Download Audio (9.0MB - audio in MP3 format). Right-click and select Save As to store the file on your local computer for later playback. Remember that you can fast forward to any question using the time index.

Related information

WebSphere Support Technical Exchange

Rate this page:

(0 users)Average rating

Add comments

Document information

More support for:

WebSphere Message Broker

Software version:


Operating system(s):

AIX, HP-UX, Linux, OS/390, Solaris, Windows, z/OS

Reference #:


Modified date:


Translate my page

Machine Translation

Content navigation