axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tim McClure <Tim.McCl...@casi-rusco.com>
Subject RE: Performance
Date Wed, 07 Nov 2001 14:47:43 GMT
Alek,

Actually we saw the the AXIS implementation was sending the HTML and XML in
different packets where XML-RPC was doing it in one.  I looked in the
org.apache.axis.transport.http.HTTPSender and saw it was doing two
consecutive socket writes (one for the HTML headers and one for the XML
payload).  When I changed the code to do one socket write instead of two the
AXIS performance increased 5-fold.  I was running my tests using a Windows
client (Linux server) and feel this may be an operating system issue

Tim

-----Original Message-----
From: Aleksander Slominski [mailto:aslom@cs.indiana.edu]
Sent: Tuesday, November 06, 2001 10:55 PM
To: axis-dev@xml.apache.org
Subject: Re: Performance


Tim McClure wrote:

> I apoligize if this is a resubmission.  I am sending it again because I
sent
> it before I was confirmed on the mailing list.
>
> I am do some out of box testing between XML-RPC and AXIS SOAP and am
> experience a significant perfromance difference.  XML-RPC processes 331
> message per second while AXIS only does 21 message per second.  I expected
a
> SOAP implementation to be slower than an XML-RPC implementation but not by
> that much.  Does anyone no why the AXIS test was so slow?  Is it possibly
a
> configuration issue ( i.e. XML validation turned on) ?

hi,

first step in direction to do any performance evaluation is to decide
exactly what
to evaluate (it seems you evaluate method invocations but what is on the
wire?)
and how to repeat evaluation (so others could run tests and compare results)
- so could you give more details about your setup, tests used and how
other people could reproduce your results?

thanks,

alek


Mime
View raw message