Re: Axis performance, additional info

[Da...@kp.org]

I converted another Apache SOAP 2.3.1 web service and test client to Axis 
1.0. This service retrieves about 75 records worth of data, versus the 
service mentioned below which returns about 5 records worth of data.

This second test seems to be more evenly matched, SOAP averaged 2556 
milliseconds per call versus 4850 milliseconds for Axis. That's about 
1.89% longer for Axis in this case. It seems the larger data set is 
leveling the playing field a bit.

- David



Jesse, 

I've noticed the speed claims as well, but my experience tells me 
otherwise. I'm not aware of any benchmarks. 

I haven't modified any of the logging defaults, so I believe it's set to 
INFO now. You are referring to logging output and the log4j.properties in 
the axis.jar, are you not? 

My test did not include any time for console output, it was just the time 
to put the call together, execute it, and get the response back. By the 
way, I'm doing RPC style calls. 

I'd welcome any other suggestions for things to check. Or pointers to 
tests that others have done. 

Thanks, 
-David 




Hi David,

I've noticed that the Axis website claims speed gains over ApacheSOAP,
but I've never really seen any evidence to back this up.  I'd be really
curious to know what kinds of benchmarks have been done comparing the
performance of the two products.

I will ask, though, whether or not debugging output could have been an
issue in your scenario.  I don't know exactly how you are using the
product, but I know that initially we had the output set to DEBUG mode
for our entire application (which included Axis) and this cost us a lot
in terms of performance.  Moving Axis to the INFO level sped things up
dramatically.

Anyway, just a few random thoughts.
-- 
=======================================
Jess Sightler
Programmer
Exim Technologies
131 Falls Street
Greenville SC 29601
Phone: 864-679-4651



On Fri, 2002-11-01 at 15:01, David.Pool@kp.org wrote:
> 
> I took a web service I created with Apache SOAP 2.3.1 and converted it
> to work with Axis 1.0. They are pretty much the same actually. I got
> curious about the performance, so I made a few simple timed tests, and
> I'm a little surprised about the results. A SOAP call took 818
> milliseconds on average, and Axis took 3415 milliseconds. In other
> words, a call via Axis was roughly 4 times longer. Ouch! I'm wondering
> if anyone else has migrated services from the old SOAP to Axis, only to
> find them much slower? 
> 
> - David
=======================================

Re: Axis performance, additional info

[Martin Jericho <ma...@jabmail.com>]

I did some tests just with axis a while ago, and found that its performance deteriorates dramatically as the message size gets larger.  I posted my results in this mailing list under the subject "Performance problems with RPC messages over 20k", but I don't think anyone has looked into it.



  ----- Original Message ----- 
  From: David.Pool@kp.org 
  To: axis-user@xml.apache.org 
  Sent: Saturday, November 02, 2002 10:39 AM
  Subject: Re: Axis performance, additional info



  I converted another Apache SOAP 2.3.1 web service and test client to Axis 1.0. This service retrieves about 75 records worth of data, versus the service mentioned below which returns about 5 records worth of data. 

  This second test seems to be more evenly matched, SOAP averaged 2556 milliseconds per call versus 4850 milliseconds for Axis. That's about 1.89% longer for Axis in this case. It seems the larger data set is leveling the playing field a bit. 

  - David 



  Jesse, 

  I've noticed the speed claims as well, but my experience tells me otherwise. I'm not aware of any benchmarks. 

  I haven't modified any of the logging defaults, so I believe it's set to INFO now. You are referring to logging output and the log4j.properties in the axis.jar, are you not? 

  My test did not include any time for console output, it was just the time to put the call together, execute it, and get the response back. By the way, I'm doing RPC style calls. 

  I'd welcome any other suggestions for things to check. Or pointers to tests that others have done. 

  Thanks, 
  -David 




  Hi David,

  I've noticed that the Axis website claims speed gains over ApacheSOAP,
  but I've never really seen any evidence to back this up.  I'd be really
  curious to know what kinds of benchmarks have been done comparing the
  performance of the two products.

  I will ask, though, whether or not debugging output could have been an
  issue in your scenario.  I don't know exactly how you are using the
  product, but I know that initially we had the output set to DEBUG mode
  for our entire application (which included Axis) and this cost us a lot
  in terms of performance.  Moving Axis to the INFO level sped things up
  dramatically.

  Anyway, just a few random thoughts.
  -- 
  =======================================
  Jess Sightler
  Programmer
  Exim Technologies
  131 Falls Street
  Greenville SC 29601
  Phone: 864-679-4651



  On Fri, 2002-11-01 at 15:01, David.Pool@kp.org wrote:
  > 
  > I took a web service I created with Apache SOAP 2.3.1 and converted it
  > to work with Axis 1.0. They are pretty much the same actually. I got
  > curious about the performance, so I made a few simple timed tests, and
  > I'm a little surprised about the results. A SOAP call took 818
  > milliseconds on average, and Axis took 3415 milliseconds. In other
  > words, a call via Axis was roughly 4 times longer. Ouch! I'm wondering
  > if anyone else has migrated services from the old SOAP to Axis, only to
  > find them much slower? 
  > 
  > - David
  =======================================