axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dennis Sosnoski <...@sosnoski.com>
Subject Re: Does Axis2 provide much better performance than Axis 1.X?
Date Wed, 18 Jun 2008 21:58:16 GMT
Having the Axis2 client communicating with an Axis 1 service operate 
twice as fast as an Axis2 client communicating with an equivalent Axis2 
service is certainly surprising. Do you know the size of the messages 
you're sending? For large messages the actual data handling (where Axis2 
should be very good) should be the most important factor in performance. 
For small messages performance is more likely to be governed by the 
actual connection handling and processing path that connects to the 
service code, and a minor glitch in the handling can have a large impact.

Also, why do you need two separate versions of the client? I'd think you 
should be able to use a common WSDL for the two implementations, so that 
you only need one client implementation in each framework and can switch 
the target URL to have it access either service implementation. That 
would at least make a better comparison.

  - Dennis

-- 
Dennis M. Sosnoski
SOA and Web Services in Java
Axis2 Training and Consulting
http://www.sosnoski.com - http://www.sosnoski.co.nz
Seattle, WA +1-425-939-0576 - Wellington, NZ +64-4-298-6117



mai.sun@nordea.com wrote:
> Hi,
>  
> Thanks for your input.
> I noticed that the output of Axis2 include prefix of namespace(s) for 
> each element while in Axis1 it uses the default namespace. This 
> actually makes response of Axis2 larger than Axis1.
> I did some more performance tests based on two web services 
> implemented with Axis 1 and Axis 2, they provide exactly the same 
> service and both of them use Wrapped style. For each deployed service 
> mention before, I generated two version of clients (Axis 1 client and 
> Axis 2 client), so in total I have 4 clients:
>  
> 1. Axis 1 client for Axis 1 service
> 2. Axis 1 client for Axis 2 service
> 3. Axis 2 client for Axis 1 service
> 4. Axis 2 client for Axis 2 service
>  
> I found that client 1 and 2 are much slower than 3 and 4 (this 
> explains Axis 2 can do XML binding faster than Axis 1). However. the 
> performance of test 3 is two times faster than test 4, which makes me 
> very curious.
>  
> Does it have something to do with the way how Axis 2 generate XML responses 
> (with namespace prefix as mentioned before)?
>  
>  
> Best wishes,
> Mai Sun 
> ------------------------------------------------------------------------
> *From:* Amila Suriarachchi [mailto:amilasuriarachchi@gmail.com]
> *Sent:* 17 June 2008 15:16
> *To:* axis-dev@ws.apache.org
> *Subject:* Re: Does Axis2 provide much better performance than Axis 1.X?
>
>
>
> On Mon, Jun 16, 2008 at 4:07 PM, <mai.sun@nordea.com 
> <mailto:mai.sun@nordea.com>> wrote:
>
>     Dear Keith,
>      
>     Thanks for your reply.
>     We're aware that Axis2 is much faster than Axis1.X and we have
>     conducted similar tests as described in the articles.
>     Our question is, is the performance gain in Axis2 is due to the
>     use of StAX API? In my opinion StAX is not faster than SAX , so
>     how come Axis2 is 5 times faster than Axis1.X?
>
>  
>
>     (We assumed the ADB binding part in Axis1.X performs as well as
>     Axis 2).
>
>
> I am not much familiar with Axis1.x ADB. But Axis2 ADB directly read 
> the underline stax Api when creating the data binding objects. if 
> Axis1.x creates any intermediate object structure then that may reduce 
> the performance.
>
> thanks,
> Amila.
>
>      
>     Best wishes,
>     Mai Sun
>
>     ------------------------------------------------------------------------
>     *From:* keith chapman [mailto:keithgchapman@gmail.com
>     <mailto:keithgchapman@gmail.com>]
>     *Sent:* 16 June 2008 11:48
>     *To:* axis-dev@ws.apache.org <mailto:axis-dev@ws.apache.org>
>     *Subject:* Re: Does Axis2 provide much better performance than
>     Axis 1.X?
>
>     Hi,
>
>     Here are some results of performance testing We've done. You can
>     go through these articles to see the reality (Axis2 is FAST).
>
>     http://wso2.org/library/91
>     http://wso2.org/library/588
>
>     Thanks,
>     Keith.
>
>     On Mon, Jun 16, 2008 at 2:43 PM, <mai.sun@nordea.com
>     <mailto:mai.sun@nordea.com>> wrote:
>
>         Hi,
>
>         Sorry for double posting…...
>
>         I read somewhere that Axis2 provides better performance
>         comparing to Axis 1.X due to the introduction of StAX XML
>         parsing. Based some tests, I can see that for large XML
>         document Axis2 is 5 to 10 times faster than Axis 1.X. I'm a
>         bit confused here since Axis1.X uses SAX parser which should
>         be at least as fast as StAX parser, so why the performance
>         improvement is achieved?
>
>         Thanks a lot!
>
>         Regards,
>         Mai Sun
>
>
>
>
>     -- 
>     Keith Chapman
>     Senior Software Engineer
>     WSO2 Inc.
>     Oxygenating the Web Service Platform.
>     http://wso2.org/
>
>     blog: http://www.keith-chapman.org
>
>
>
>
> -- 
> Amila Suriarachchi,
> WSO2 Inc. 

---------------------------------------------------------------------
To unsubscribe, e-mail: axis-dev-unsubscribe@ws.apache.org
For additional commands, e-mail: axis-dev-help@ws.apache.org


Mime
View raw message