Tuning AquaLogic Service Bus

Whenever possible, set the logging level to warning. You set the logging level in the WebLogic Server Administration Console. For more information, see Servers: Logging: General in the WebLogic Server Administration Console Online Help. The following code displays the output server config.xml file when the logging level is set to warning. For more information on logging, see “Log” in Proxy Services: Actions in the Using the AquaLogic Service Bus Console.

<server>

  <name>AdminServer</name>

  <log>

    <file-min-size>5000</file-min-size>

    <log-file-severity>Warning</log-file-severity>

    <log-file-filter xsi:nil="true"></log-file-filter>

    <stdout-severity>Off</stdout-severity>

    <stdout-filter xsi:nil="true"></stdout-filter>

    <domain-log-broadcast-severity>Error</domain-log-broadcast-severity>

    <domain-log-broadcast-filter xsi:nil="true"></domain-log-broadcast-filter>

    <memory-buffer-severity>Error</memory-buffer-severity>

    <memory-buffer-filter xsi:nil="true"></memory-buffer-filter>

  </log>

</server>

Group JMS queues on different JMS servers based on message loads. Different JMS servers use different file stores, which you can distribute to separate disk volumes. For more information, “Adding a Business Service” in Business Services in the Using the AquaLogic Service Bus Console and pay particular attention to the JMS configuration information.
If you are using an Oracle database as a JMS persistent store, it is recommended that you use a 10g database and ensure that it has sufficient JDBC connections. Create a JDBC store on a separate schema to use a separate tablespace.
If you do not require monitoring for a proxy or business service, disable the monitoring capability. For more information, see “Overview of Monitoring” in Monitoring in the Using the AquaLogic Service Bus Console.
If possible, set the routing data in the JMS message properties. AquaLogic Service Bus does not deserialize message content until the content is explicitly accessed in the pipeline. For example, if the content is an XML document, XML parsing does not happen until an XQuery or XSLT operation happens in the pipeline. For more information about working with the message context in the message flow, see Message Context.
If you need to extract some of the inbound header elements for processing, you should specify that AquaLogic Service Bus retrieves specific header elements instead of all the elements.
If you are using an Oracle database as a JMS persistent store BEA recommends that you should ensure that enough JDBC connections are available. This allows an administrator to tune the block size based upon the message size. It is also possible to create each tablespace on a separate datafile and put these datafiles on separate disks. Such an arrangement allows for greater concurrency at the hardware level when the datafiles are not stored on a RAID.
Where possible, use the insert action instead of the assign action. The insert action uses “in-place” modification semantics enhancing the performance compared to the assign action. For information on configuring actions, see “Adding an Action” in Proxy Services: Actions in the Using the AquaLogic Service Bus Console.
Use AquaLogic Service Bus clustering and WebSphere MQ clustering to achieve scalability.
When you are configuring the Accept Backlog parameter in the WebLogic Server BEA recommends that you should first increase the default value by twenty five percent.If the ‘connection refused, socket exception’ is thrown, then continue increasing the value till the exception is not thrown. For more information on tuning the WebLogic Server see Tuning WebLogic Server.
If a front-end application invokes AquaLogic Service Bus synchronously, AquaLogic Service Bus can use the synchronous-asynchronous feature to communicate with WebSphere MQ synchronously. On the WebSphere MQ side, a request and a response queue is set up. AquaLogic Service Bus sends a request to the request queue and waits for a response from the response queue. To achieve improved performance, you can use a dedicate work manager for the response Message Driven Bean. You configure the dedicate work manager in the WebLogic Server Administration Console. For more information, see Work Manager in the WebLogic Server Administration Console Online Help. The following code displays the output server config.xml file after the dedicate work manager is configured.

<self-tuning>

 <min-threads-constraint>

    <name>minThreadsConstraint</name>

    <target>AdminServer</target>

    <count>20</count>

 </min-threads-constraint>

 <work-manager>

    <name>MQWorkManager</name>

    <target>AdminServer</target>

    <min-threads-constraint> minThreadsConstraint </min-threads-constraint>

    <ignore-stuck-threads>false</ignore-stuck-threads>

 </work-manager>

</self-tuning>

AquaLogic Service Bus uses a two-level cache for proxy service run-time data with static and dynamic sections. The static section is a size-based Least Recently Used (LRU) cache for proxy services, immune from garbage collection. When a proxy service is bumped from the static section, it is demoted to the dynamic section. Proxy services in the dynamic section can continue to be reused but are susceptible to being deleted during a Java garbage collection cycle.

The cache introduces a performance trade off between memory consumption and compilation cost. Caching proxy services increases throughput but consume memory otherwise available for data processing.

You can tune the cache by setting its size using the system property com.bea.wli.sb.pipeline.RouterRuntimeCache.size

The default value is 100. This is the number of proxy services in the static portion of the cache.

Approximately every minute, a log message will be printed with the following format:

[RuntimeRouterCache] XX hits received: YY hits to main cache, ZZ hits to soft cache, MM misses

Where:

XX represents the number of times that the cache was accessed since the last log.
YY represents the nummer of times (out of XX) that the cache contained the proxy service.
ZZ represents the number of times (out of XX) that the cache contained the proxy service amongst those to be garbaged collected (for example, the cache was able to reclaim the proxy service before the garbage collector did).
MM represents the number of times (out of XX) that the cache did not contain the proxy service and had to recreate it.

You can use these logs to tune the size of the cache. Consider the following guidelines:

If your AquaLogic Service Bus domain contains a low number of proxy services, BEA recommends setting the size of the cache greater than that number.
If your AquaLogic Service Bus domain contains a large number of proxy services, adjust the size of the cache to decrease the percentage of missed hits and still allow enough memory for data processing.

User Guide

Tuning AquaLogic Service Bus