cocoon-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Stefano Mazzocchi <>
Subject Re: [FYI] Profiling Cocoon...
Date Sun, 06 Oct 2002 20:11:12 GMT
[back to the list since others might be interested in this]

Ignacio J. Ortega wrote:

> Can you elaborate a little more on this point? do you have any numbers
> to support the relative badness of this 10%?

Hmmm, I don't know what numbers I can give you since any absolute number 
will probably be not meaningful on an instrumented JVM.

Anyway, the thing is that, on average, 10% of the time spent by the 
thread reading the http request, processing and writing the response is 
10% of the total time spent (not talking about java's own IO times, this 
is just catalina doing stuff)

Now, if we suppose that these numbers are not influenced by the 
instrumentation of the JVM (which is a reasonable estimation since 
Catalina is pure java), a 1000ms response has catalina spending 100ms 
doing stuff.

But this was tested with the HTTP connector. We didn't test it with the 
AJP connectors that leave all the HTTP parsing to Apache HTTPD (but 
might add some over-the-wire costs)

Anyway, it seems to me kind of absurd that Cocoon does so much stuff in 
70% of the response time, while 20% is spent on IO and 10% is spent by 
Catalina to call CocoonServlet.service(). That's all.

Stefano Mazzocchi                               <>

To unsubscribe, e-mail:
For additional commands, email:

View raw message