In researching problems with SOAP::Lite, I hit upon this PDF (Google HTML version).

I was quite surprised to see how much slower SOAP::Lite was compared to other implementations - it's latency is 35 times that of the JavaRMI implementation, with actual requests being about 100 times slower (in the examples they used).

Not knowing much about TCP, I'm not exactly sure what the problem is as described, but one comment stands out:

"SOAP::Lite automatically serializes associative arrays where the same object may be referenced more than once. My guess is that while this is very convenient, it has a large impact on performance."

Is anyone else aware of these issues? How easy would the problems be to address? We've decided to move away from Perl on the server end of transactions and will wait and see what happens. If you use SOAP, what approach have you taken?

Thoughts anyone?

cLive ;-)

Edit: OK, I misread it, but I'm still interested in whether anyone's been looking at speeding up the module :)

  • Comment on SOAP::Lite - worst SOAP implementation?

Replies are listed 'Best First'.
Re: SOAP::Lite - worst SOAP implementation?
by perrin (Chancellor) on Apr 12, 2005 at 19:04 UTC
    Hang on there. The Java RMI being measured there is not SOAP. It's just there to show you how much SOAP sucks compared to basically any of the other remote method protocols that have been around for years.

    Now, granted, SOAP::Lite is still twice as slow as the SOAP RMI in Java, but that's practically line-noise when you consider how monumentally slow SOAP is at doing much of anything.

      "The Java RMI being measured there is not SOAP."

      /me goes back to actually read the article in detail :)

      The Apache Axis server is more than twice as fast as SOAP::Lite - especially on some of the larger tests (from what I can see in my second quick read-through. Hmmmmm, headache time. I must be turning into management. I glaze over on the details a lot quicker these days ;-)

      cLive ;-)

        I must be turning into management. I glaze over on the details a lot quicker these days ;-)

        No, you are not management. The difference between management and just growing older is that management wouldn't bother to re-read the article. :-)

Re: SOAP::Lite - worst SOAP implementation?
by samtregar (Abbot) on Apr 12, 2005 at 20:41 UTC
    If you use SOAP, what approach have you taken?

    I've used SOAP a few times, but it's usually been more trouble than its worth. Recently I decided to factor SOAP out of a new design and the results have been pretty good. Check it out:

    http://www.perl.com/pub/a/2004/09/30/drop_the_soap.html

    -sam