For simple things on my machine, I get upwards of 5K or more per sec.   Do you 
have any HTTP filters or anything that may be causing an issue?

I would start with something simple like CXF's wsdl_first example, try running 
that, and see what you get.   Then start adding in some of your configs and 
such.


Dan




On Sep 11, 2012, at 1:08 PM, rouble <[email protected]> wrote:

> CXF Gurus,
> 
> We are using CXF for a bunch of web services on our enterprise grade
> servers. We were noticing that our throughput of requests was peaking
> at an average of 42 transactions/second (a transaction here is 1 SOAP
> request/response combo).
> 
> To investigate whether the bottleneck was in the system or in the web
> service, we added a simple API (getStartTime) that just returns a
> string without communicating with the rest of the system. The API has
> no parameters, and just returns a small string (timestamp) - so the
> size of the message is very small.
> 
> We used SOAPUI's load test feature to run multiple threads against the
> server that just call getStartTime. However, once again, we can only
> get a max throughput of 42 transactions per second - no matter how
> many client threads we have. The system CPU is about 50% idle during
> the test and we do not seem to be memory bound either.
> 
> We're running CXF 2.4.3 in Tomcat 7.0. Tomcat's config had
> acceptCount="100", maxThreads="150" - which should not cause any
> throttling to kick in. I set both those to 300 but that did not change
> our throughput. I checked iptables - there is no throttling going on
> there either. Since I have checked everything else it seems like the
> bottleneck is in CXF. Is there something in CXF that throttles
> requests? If so, is it configurable?
> 
> I know 42 is the answer to life the universe and everything but does
> anyone have other thoughts on how to increase our system throughput?
> 
> tia,
> rouble

-- 
Daniel Kulp
[email protected] - http://dankulp.com/blog
Talend Community Coder - http://coders.talend.com

Reply via email to