[jira] [Updated] (AXIS2-5118) In high load scenarios with many threads, the call to Introspector.getBeanInfo(Class,Class) causes high synchronization resulting in system stall

["Ronald Brindl (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/AXIS2-5118?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Ronald Brindl updated AXIS2-5118:
---------------------------------

    Attachment: BeanInfoCacheTest.java
                BeanUtil.java
                BeanInfoCache.java

Added BeanInfoCache.java, which caches BeanInfo Objects using a limited linked list, which removes oldest entries based on maximum size (currently MAX_SIZE = 10). 
It is thread safe using a ReentrandReadWrite Lock for accessing its backing list.

In BeanInfo i replaced the call
{code}
BeanInfo beanInfo = Introspector.getBeanInfo(beanClass, beanClass.getSuperclass());
{code}
with 
{code}
            BeanInfo beanInfo = BeanInfoCache.get(beanClass);
            if(beanInfo == null) {
            	beanInfo = Introspector.getBeanInfo(beanClass, beanClass.getSuperclass());
            	BeanInfoCache.put(beanClass, beanInfo);
            }
{code}


> In high load scenarios with many threads, the call to Introspector.getBeanInfo(Class,Class) causes high synchronization resulting in system stall
> -------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: AXIS2-5118
>                 URL: https://issues.apache.org/jira/browse/AXIS2-5118
>             Project: Axis2
>          Issue Type: Improvement
>          Components: adb
>    Affects Versions: 1.5.4
>         Environment: Tomcat 7 with axis2 webservices
>            Reporter: Ronald Brindl
>         Attachments: BeanInfoCache.java, BeanInfoCacheTest.java, BeanUtil.java
>
>
> I have a high load scenario, where a Clustered Web frontend does Webservice calls to a Backend Webservices. 
> There are around 200 Threads handling the webservices.
> The CPU load of the machine went to near zero.
> Using dynatrace, we realized, that all the calls went through BeanUtils.getPropertyQnameList, which in turn calls Introspector.getBeanInfo(beanClass, beanClass.getSuperclass());
> This again calls WebAppClassLoader.loadClass(String name, boolean resolve), which is synchronized.
> I fixed this problem by modifying BeanUtil.getPropertyQnameList so that it caches BeanInfo objects.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
For additional commands, e-mail: java-dev-help@axis.apache.org