Re: SOAP performance data and open benchmarks

We developed the gSOAP toolkit for C and C++ at Florida State University and
tested several SOAP packages to compare ease-of-use and performance. The
investigation is part of a federally funded research project. The tests were
subject to change as SOAP toolkits were maturing and so we didn't publish
the results of this investigation yet.

IMO, the better toolkits are easy to use (e.g. fully automated and offer
more than just a library-based API) and provide a fast and scalable
solution.

Through research (and using common sense), it is clear that it is not always
possible to achieve equal or better performance than Java RMI, because
SOAP/XML is bandwidth demanding. Sure, if you discount communication
latencies than SOAP/XML could be just as fast compared to Java RMI. However,
certain types of messages such as arrays of floats introduce a significant
overhead in communication. Messages that exchange strings, base64, and DIME
attachments are less demanding and performance is close to Java RMI.
Therefore, an unbiased performance comparison should consider differrent
types of messages, data, and encodings.

A secondary issue is the overhead introduced with CGI, Fast-CGI, or Apache
mod/IIS modules. Preferably, a Web service should be a stand-alone.

Here is a paper that compares performance and scalabolity of CGI vs.
stand-alone vs. Java RMI and compares the speed of different encodings:
http://www.cs.fsu.edu/~engelen/soappaper.html

The following paper focusses on performance and presents the best possible
(de)serialization speeds that a tookit can achieve:
http://www.extreme.indiana.edu/xgws/papers/soap-hpdc2002/soap-hpdc2002.pdf
We commented on these results at:
http://www.cs.fsu.edu/~engelen/soapperf.html

- Regards, Robert van Engelen, FSU Computer Science.

You can read messages from the SOAP archive, unsubscribe from SOAP, or subscribe to other
DevelopMentor lists at http://discuss.develop.com.