cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Daniel Fagerstrom <dani...@nada.kth.se>
Subject Re: Performance with blocks protocol
Date Wed, 08 Nov 2006 10:12:00 GMT
Lars Trieloff skrev:
> Results from a JMeter-drilling test: accessing 
> http://localhost:8080/cocoon/cocoon-blocks-fw-sample3/sub/test in 10 
> parallel threads using the cocoon-blocks-fw-sample.
>
> - 1 DispatcherServlet
> - 4 BlockServlet
> - 4 SitemapServlet
>
> - 6 TreeProcessor
> - 6 EnvironmentHelper
Seem like the singleton settings work this far. A little bit unclear why 
there are two extra TreeProcessors though. One of them might be from a 
"standard" SitemapServlet.

> - 2731 BlockCallHttpServletRequest
>
> - 1 ReadNode
>
> - 2731 ResourceReader
>
> - 5463 HttpEnvironment
> - 5463 HttpRequest
> - 5463 HttpResponse
> - 5463 NonCachingProcessingPipeline
Here it seem like something is wrong with cleanup after sitemap 
processing. The SitemapServlet creates a new HttpEnvironment object that 
refer to the incoming request and response object at each request. The 
environment object is in turn referenced by the 
NonCachingProcessingPipeline object which also is responsible for th 
ResourceReader object.

The problem seem to be that the recycle method of the pipeline is not 
called and that the pooling doesn't work. It is not that easy to follow 
the code, so I don't yet know exactly what happens or where.

Do you have similar figures from Cocoon used without the blocks fw? It 
would help tracking down the problem to know if it is block specific or not.

/Daniel


Mime
View raw message