[wildfly]jmx call performance degradates a lot from JBoss42 to Wildfly9

[lcheng_nj]

We switch our application server from  JBoss4.2 to wildfly9.0.1 recently. After some testing, we found the JMX call performance degrade a lot.

Take following test method as example, we use it to send one string array, the array size can vary, and each string length can change.

public void perfWithStringArray(String[] events)

It's easy to reproduce in my environment. Start application server (JBoss42 or Wildfly9) in my machine, and then deploy one MBean(notifiableMBeanjboss42.sar for JBoss42, notifiableMBeanwildfly.jar for wildfly) with following simple method.

public void perfWithStringArray(String[] events) .

The client (StandaloneMBeanInvoker4JBoss42.java for JBoss42, and StandaloneInvoker.java for wildfly9) is also located in my local machine. Please take a look at the performance data

StringLength*ArrayLength	EPS(JBoss42)	EPS(Wildfly9.01)
128*200	300000	65000
128*500	365000	80000
512*500	126000	25000
1024*500	67000	12000
1024*5000	67000	12000

The last two columns mean how many event could be handled in one second. You can see the performance degrades a lot.

Not sure whether wildfly has any setting/configuration that could be modified/tuned to improve the performance.

The MBean and client code is attached.

[dlofthouse]

I would suggest creating an issue within the REMJMX project with as much information as you can provide, we would need to work out if this is actually Remoting or the Remoting JMX protocol - initially I would imagine we have some options in the latter to optimise it further.

[jaikiran]

In addition to what Darran suggested, can you also get the numbers against the latest released WildFly 11.0.0.Final?

[lcheng_nj]

Thanks. Currently, do we have any options that I can configure in configuration file. e.g. standalone-full.xml

[lcheng_nj]

Sorry, I don't think we will try Wildfly 11 recently.

[lcheng_nj]

I filed one bug to Remoting component, please have a look.

https://issues.jboss.org/browse/REM3-317

[lcheng_nj]

I update the description and performance data, and I also update the information in https://issues.jboss.org/browse/REM3-317 .

It's very easy to reproduce. That is, deploy the mbean in local application server and then run the client in the local machine two. Since the performance degrades two much, please give some help/comment.