You are viewing a plain text version of this content. The canonical link for it is here.
Posted to java-dev@axis.apache.org by Charith Wickramarachchi <ch...@gmail.com> on 2011/11/29 12:21:03 UTC
Improving JAX-WS Performance
Hi Devs ,
JAX-WS is becoming a popular key webservice development style. While
supporting its own web-service development/deployment style i think Axis2
should improve its JAX-WS implementation. I started to profile Axis2 JAX-WS
implementation having performance improvement in mind and was able to
isolate a major bottleneck in this flow. (See the attached profile image.)
Here the bottleneck was AXOM creating a XMLOutput factory per each
invocation due to a bug in JAXWSMessageReceiver. I submitted a fix for this
in[1]. And we are currently in the process of discussing on a proper
solution for that issue.
With this patch i was able to get ~75% performance improvement on Simple
Echo jax-ws services and ~60% improvement with services with complex types.
With this observation i think if we spend some energy on improving Axis2
JAX-WS performance we will be able to improve it a lot and
if we can publish that improved numbers may be with comparisons that will
add a great value to this project.
So i though of starting a discussion around it to discuss on how to improve
this and areas we must focus on.
[1]https://issues.apache.org/jira/browse/AXIS2-5203
thanks,
Charith
--
Charith Dhanushka Wickramarachchi
http://charithwiki.blogspot.com/
Re: Improving JAX-WS Performance
Posted by Sagara Gunathunga <sa...@gmail.com>.
I tried after adding @RequestWrapper to the service method and could see
those data used in deployment time too but in runtime there is no
difference in processing time. I noticed now it builds message in a
different method. I have limited knowledge on these codes but as I read
this seems an issue with MethodMarshaller classes ( see [1] ), the way they
process incoming messages.
IMO since this is the most simplest form of JAX-WS service we have to fix
this.
[1] -
https://svn.apache.org/repos/asf/axis/axis2/java/core/trunk/modules/jaxws/src/org/apache/axis2/jaxws/marshaller/impl/alt/DocLitWrappedMinimalMethodMarshaller.java
Thanks !
On Sat, Dec 3, 2011 at 6:27 AM, Andreas Veithen
<an...@gmail.com>wrote:
> I think that the issue only occurs with services that use parameter
> style WRAPPED and that have no @RequestWrapper annotation (or that
> have a @RequestWrapper annotation but that don't package the wrapper
> beans). I think that for WSDL-first, request wrappers are generated
> automatically, so this would only apply to services implemented using
> the code-first approach. Nevertheless this still covers a wide range
> of use cases, so that this is a problem.
>
> Andreas
>
> On Sat, Dec 3, 2011 at 01:26, Sagara Gunathunga
> <sa...@gmail.com> wrote:
> > No.
> >
> > This is my method signature.
> > @WebService
> > @SOAPBinding(style = Style.DOCUMENT, use=Use.LITERAL)
> > public interface OrderService {
> >
> > @WebMethod
> > Order placeOrder(@WebParam(name="order")Order order);
> >
> > }
> >
> > Thanks !
> >
> > On Fri, Dec 2, 2011 at 3:55 PM, Andreas Veithen <
> andreas.veithen@gmail.com>
> > wrote:
> >>
> >> Do your service methods have @RequestWrapper annotations?
> >>
> >> Andreas
> >>
> >> On Sat, Dec 3, 2011 at 00:31, Sagara Gunathunga
> >> <sa...@gmail.com> wrote:
> >> > I profiled a complex JAX-WS web service (refer the attachment) it
> seems
> >> > 45%
> >> > of time taken for JAXB unmarshalling while another 28.9% of time
> taken
> >> > for
> >> > message.getBodyBlockQNames() that eventually build the message through
> >> > AXIOM. I experimented hard coding some of the value calculation to
> avoid
> >> > this message building but what I realized is it's not a single place
> >> > JAX-WS
> >> > build this message instead when i avoid message building in a one
> place
> >> > it
> >> > build the message in a later stage.
> >> >
> >> > I can't realize why we have to build the message instead of passing it
> >> > to
> >> > JAXB Unmarshaller without building the message.
> >> >
> >> >
> >> > For me this seems like a extra cost , can any one explain the rational
> >> > behind this idea ? or can't we git rid of this message building.
> >> >
> >> > Thanks !
> >> >
> >> > On Wed, Nov 30, 2011 at 5:27 AM, Isuru Suriarachchi <
> isurues@gmail.com>
> >> > wrote:
> >> >>
> >> >> Hi Charith,
> >> >>
> >> >> +1 for putting some more effort on improving our JAX-WS support. I'll
> >> >> commit the proper fix to [1] ASAP and that will remove the bottleneck
> >> >> in the
> >> >> JAX-WS MR. And also there'll be more places at which we can do
> >> >> improvements
> >> >> to gain even better performance numbers.
> >> >>
> >> >> Thanks,
> >> >> ~Isuru
> >> >>
> >> >> [1] https://issues.apache.org/jira/browse/AXIS2-4930
> >> >>
> >> >> On Tue, Nov 29, 2011 at 4:51 PM, Charith Wickramarachchi
> >> >> <ch...@gmail.com> wrote:
> >> >>>
> >> >>> Hi Devs ,
> >> >>>
> >> >>> JAX-WS is becoming a popular key webservice development style.
> While
> >> >>> supporting its own web-service development/deployment style i think
> >> >>> Axis2
> >> >>> should improve its JAX-WS implementation. I started to profile Axis2
> >> >>> JAX-WS
> >> >>> implementation having performance improvement in mind and was able
> to
> >> >>> isolate a major bottleneck in this flow. (See the attached profile
> >> >>> image.)
> >> >>> Here the bottleneck was AXOM creating a XMLOutput factory per each
> >> >>> invocation due to a bug in JAXWSMessageReceiver. I submitted a fix
> for
> >> >>> this
> >> >>> in[1]. And we are currently in the process of discussing on a proper
> >> >>> solution for that issue.
> >> >>>
> >> >>> With this patch i was able to get ~75% performance improvement on
> >> >>> Simple
> >> >>> Echo jax-ws services and ~60% improvement with services with complex
> >> >>> types.
> >> >>>
> >> >>>
> >> >>> With this observation i think if we spend some energy on improving
> >> >>> Axis2
> >> >>> JAX-WS performance we will be able to improve it a lot and
> >> >>> if we can publish that improved numbers may be with comparisons that
> >> >>> will
> >> >>> add a great value to this project.
> >> >>>
> >> >>> So i though of starting a discussion around it to discuss on how to
> >> >>> improve this and areas we must focus on.
> >> >>>
> >> >>>
> >> >>> [1]https://issues.apache.org/jira/browse/AXIS2-5203
> >> >>>
> >> >>> thanks,
> >> >>> Charith
> >> >>>
> >> >>> --
> >> >>> Charith Dhanushka Wickramarachchi
> >> >>> http://charithwiki.blogspot.com/
> >> >>>
> >> >>>
> >> >>>
> >> >>>
> ---------------------------------------------------------------------
> >> >>> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> >> >>> For additional commands, e-mail: java-dev-help@axis.apache.org
> >> >>
> >> >>
> >> >>
> >> >>
> >> >> --
> >> >> Technical Lead,
> >> >> WSO2 Inc. http://wso2.org/
> >> >> Blog : http://isurues.wordpress.com/
> >> >
> >> >
> >> >
> >> >
> >> > --
> >> > Sagara Gunathunga
> >> >
> >> > Blog - http://ssagara.blogspot.com
> >> > Web - http://people.apache.org/~sagara/
> >> > LinkedIn - http://www.linkedin.com/in/ssagara
> >> >
> >> >
> >> > ---------------------------------------------------------------------
> >> > To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> >> > For additional commands, e-mail: java-dev-help@axis.apache.org
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> >> For additional commands, e-mail: java-dev-help@axis.apache.org
> >>
> >
> >
> >
> > --
> > Sagara Gunathunga
> >
> > Blog - http://ssagara.blogspot.com
> > Web - http://people.apache.org/~sagara/
> > LinkedIn - http://www.linkedin.com/in/ssagara
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> For additional commands, e-mail: java-dev-help@axis.apache.org
>
>
--
Sagara Gunathunga
Blog - http://ssagara.blogspot.com
Web - http://people.apache.org/~sagara/
LinkedIn - http://www.linkedin.com/in/ssagara
Re: Improving JAX-WS Performance
Posted by Andreas Veithen <an...@gmail.com>.
I think that the issue only occurs with services that use parameter
style WRAPPED and that have no @RequestWrapper annotation (or that
have a @RequestWrapper annotation but that don't package the wrapper
beans). I think that for WSDL-first, request wrappers are generated
automatically, so this would only apply to services implemented using
the code-first approach. Nevertheless this still covers a wide range
of use cases, so that this is a problem.
Andreas
On Sat, Dec 3, 2011 at 01:26, Sagara Gunathunga
<sa...@gmail.com> wrote:
> No.
>
> This is my method signature.
> @WebService
> @SOAPBinding(style = Style.DOCUMENT, use=Use.LITERAL)
> public interface OrderService {
>
> @WebMethod
> Order placeOrder(@WebParam(name="order")Order order);
>
> }
>
> Thanks !
>
> On Fri, Dec 2, 2011 at 3:55 PM, Andreas Veithen <an...@gmail.com>
> wrote:
>>
>> Do your service methods have @RequestWrapper annotations?
>>
>> Andreas
>>
>> On Sat, Dec 3, 2011 at 00:31, Sagara Gunathunga
>> <sa...@gmail.com> wrote:
>> > I profiled a complex JAX-WS web service (refer the attachment) it seems
>> > 45%
>> > of time taken for JAXB unmarshalling while another 28.9% of time taken
>> > for
>> > message.getBodyBlockQNames() that eventually build the message through
>> > AXIOM. I experimented hard coding some of the value calculation to avoid
>> > this message building but what I realized is it's not a single place
>> > JAX-WS
>> > build this message instead when i avoid message building in a one place
>> > it
>> > build the message in a later stage.
>> >
>> > I can't realize why we have to build the message instead of passing it
>> > to
>> > JAXB Unmarshaller without building the message.
>> >
>> >
>> > For me this seems like a extra cost , can any one explain the rational
>> > behind this idea ? or can't we git rid of this message building.
>> >
>> > Thanks !
>> >
>> > On Wed, Nov 30, 2011 at 5:27 AM, Isuru Suriarachchi <is...@gmail.com>
>> > wrote:
>> >>
>> >> Hi Charith,
>> >>
>> >> +1 for putting some more effort on improving our JAX-WS support. I'll
>> >> commit the proper fix to [1] ASAP and that will remove the bottleneck
>> >> in the
>> >> JAX-WS MR. And also there'll be more places at which we can do
>> >> improvements
>> >> to gain even better performance numbers.
>> >>
>> >> Thanks,
>> >> ~Isuru
>> >>
>> >> [1] https://issues.apache.org/jira/browse/AXIS2-4930
>> >>
>> >> On Tue, Nov 29, 2011 at 4:51 PM, Charith Wickramarachchi
>> >> <ch...@gmail.com> wrote:
>> >>>
>> >>> Hi Devs ,
>> >>>
>> >>> JAX-WS is becoming a popular key webservice development style. While
>> >>> supporting its own web-service development/deployment style i think
>> >>> Axis2
>> >>> should improve its JAX-WS implementation. I started to profile Axis2
>> >>> JAX-WS
>> >>> implementation having performance improvement in mind and was able to
>> >>> isolate a major bottleneck in this flow. (See the attached profile
>> >>> image.)
>> >>> Here the bottleneck was AXOM creating a XMLOutput factory per each
>> >>> invocation due to a bug in JAXWSMessageReceiver. I submitted a fix for
>> >>> this
>> >>> in[1]. And we are currently in the process of discussing on a proper
>> >>> solution for that issue.
>> >>>
>> >>> With this patch i was able to get ~75% performance improvement on
>> >>> Simple
>> >>> Echo jax-ws services and ~60% improvement with services with complex
>> >>> types.
>> >>>
>> >>>
>> >>> With this observation i think if we spend some energy on improving
>> >>> Axis2
>> >>> JAX-WS performance we will be able to improve it a lot and
>> >>> if we can publish that improved numbers may be with comparisons that
>> >>> will
>> >>> add a great value to this project.
>> >>>
>> >>> So i though of starting a discussion around it to discuss on how to
>> >>> improve this and areas we must focus on.
>> >>>
>> >>>
>> >>> [1]https://issues.apache.org/jira/browse/AXIS2-5203
>> >>>
>> >>> thanks,
>> >>> Charith
>> >>>
>> >>> --
>> >>> Charith Dhanushka Wickramarachchi
>> >>> http://charithwiki.blogspot.com/
>> >>>
>> >>>
>> >>>
>> >>> ---------------------------------------------------------------------
>> >>> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
>> >>> For additional commands, e-mail: java-dev-help@axis.apache.org
>> >>
>> >>
>> >>
>> >>
>> >> --
>> >> Technical Lead,
>> >> WSO2 Inc. http://wso2.org/
>> >> Blog : http://isurues.wordpress.com/
>> >
>> >
>> >
>> >
>> > --
>> > Sagara Gunathunga
>> >
>> > Blog - http://ssagara.blogspot.com
>> > Web - http://people.apache.org/~sagara/
>> > LinkedIn - http://www.linkedin.com/in/ssagara
>> >
>> >
>> > ---------------------------------------------------------------------
>> > To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
>> > For additional commands, e-mail: java-dev-help@axis.apache.org
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
>> For additional commands, e-mail: java-dev-help@axis.apache.org
>>
>
>
>
> --
> Sagara Gunathunga
>
> Blog - http://ssagara.blogspot.com
> Web - http://people.apache.org/~sagara/
> LinkedIn - http://www.linkedin.com/in/ssagara
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
For additional commands, e-mail: java-dev-help@axis.apache.org
Re: Improving JAX-WS Performance
Posted by Sagara Gunathunga <sa...@gmail.com>.
No.
This is my method signature.
@WebService
@SOAPBinding(style = Style.DOCUMENT, use=Use.LITERAL)
public interface OrderService {
@WebMethod
Order placeOrder(@WebParam(name="order")Order order);
}
Thanks !
On Fri, Dec 2, 2011 at 3:55 PM, Andreas Veithen
<an...@gmail.com>wrote:
> Do your service methods have @RequestWrapper annotations?
>
> Andreas
>
> On Sat, Dec 3, 2011 at 00:31, Sagara Gunathunga
> <sa...@gmail.com> wrote:
> > I profiled a complex JAX-WS web service (refer the attachment) it seems
> 45%
> > of time taken for JAXB unmarshalling while another 28.9% of time taken
> for
> > message.getBodyBlockQNames() that eventually build the message through
> > AXIOM. I experimented hard coding some of the value calculation to avoid
> > this message building but what I realized is it's not a single place
> JAX-WS
> > build this message instead when i avoid message building in a one place
> it
> > build the message in a later stage.
> >
> > I can't realize why we have to build the message instead of passing it to
> > JAXB Unmarshaller without building the message.
> >
> >
> > For me this seems like a extra cost , can any one explain the rational
> > behind this idea ? or can't we git rid of this message building.
> >
> > Thanks !
> >
> > On Wed, Nov 30, 2011 at 5:27 AM, Isuru Suriarachchi <is...@gmail.com>
> > wrote:
> >>
> >> Hi Charith,
> >>
> >> +1 for putting some more effort on improving our JAX-WS support. I'll
> >> commit the proper fix to [1] ASAP and that will remove the bottleneck
> in the
> >> JAX-WS MR. And also there'll be more places at which we can do
> improvements
> >> to gain even better performance numbers.
> >>
> >> Thanks,
> >> ~Isuru
> >>
> >> [1] https://issues.apache.org/jira/browse/AXIS2-4930
> >>
> >> On Tue, Nov 29, 2011 at 4:51 PM, Charith Wickramarachchi
> >> <ch...@gmail.com> wrote:
> >>>
> >>> Hi Devs ,
> >>>
> >>> JAX-WS is becoming a popular key webservice development style. While
> >>> supporting its own web-service development/deployment style i think
> Axis2
> >>> should improve its JAX-WS implementation. I started to profile Axis2
> JAX-WS
> >>> implementation having performance improvement in mind and was able to
> >>> isolate a major bottleneck in this flow. (See the attached profile
> image.)
> >>> Here the bottleneck was AXOM creating a XMLOutput factory per each
> >>> invocation due to a bug in JAXWSMessageReceiver. I submitted a fix for
> this
> >>> in[1]. And we are currently in the process of discussing on a proper
> >>> solution for that issue.
> >>>
> >>> With this patch i was able to get ~75% performance improvement on
> Simple
> >>> Echo jax-ws services and ~60% improvement with services with complex
> types.
> >>>
> >>>
> >>> With this observation i think if we spend some energy on improving
> Axis2
> >>> JAX-WS performance we will be able to improve it a lot and
> >>> if we can publish that improved numbers may be with comparisons that
> will
> >>> add a great value to this project.
> >>>
> >>> So i though of starting a discussion around it to discuss on how to
> >>> improve this and areas we must focus on.
> >>>
> >>>
> >>> [1]https://issues.apache.org/jira/browse/AXIS2-5203
> >>>
> >>> thanks,
> >>> Charith
> >>>
> >>> --
> >>> Charith Dhanushka Wickramarachchi
> >>> http://charithwiki.blogspot.com/
> >>>
> >>>
> >>>
> >>> ---------------------------------------------------------------------
> >>> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> >>> For additional commands, e-mail: java-dev-help@axis.apache.org
> >>
> >>
> >>
> >>
> >> --
> >> Technical Lead,
> >> WSO2 Inc. http://wso2.org/
> >> Blog : http://isurues.wordpress.com/
> >
> >
> >
> >
> > --
> > Sagara Gunathunga
> >
> > Blog - http://ssagara.blogspot.com
> > Web - http://people.apache.org/~sagara/
> > LinkedIn - http://www.linkedin.com/in/ssagara
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> > For additional commands, e-mail: java-dev-help@axis.apache.org
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> For additional commands, e-mail: java-dev-help@axis.apache.org
>
>
--
Sagara Gunathunga
Blog - http://ssagara.blogspot.com
Web - http://people.apache.org/~sagara/
LinkedIn - http://www.linkedin.com/in/ssagara
Re: Improving JAX-WS Performance
Posted by Andreas Veithen <an...@gmail.com>.
Do your service methods have @RequestWrapper annotations?
Andreas
On Sat, Dec 3, 2011 at 00:31, Sagara Gunathunga
<sa...@gmail.com> wrote:
> I profiled a complex JAX-WS web service (refer the attachment) it seems 45%
> of time taken for JAXB unmarshalling while another 28.9% of time taken for
> message.getBodyBlockQNames() that eventually build the message through
> AXIOM. I experimented hard coding some of the value calculation to avoid
> this message building but what I realized is it's not a single place JAX-WS
> build this message instead when i avoid message building in a one place it
> build the message in a later stage.
>
> I can't realize why we have to build the message instead of passing it to
> JAXB Unmarshaller without building the message.
>
>
> For me this seems like a extra cost , can any one explain the rational
> behind this idea ? or can't we git rid of this message building.
>
> Thanks !
>
> On Wed, Nov 30, 2011 at 5:27 AM, Isuru Suriarachchi <is...@gmail.com>
> wrote:
>>
>> Hi Charith,
>>
>> +1 for putting some more effort on improving our JAX-WS support. I'll
>> commit the proper fix to [1] ASAP and that will remove the bottleneck in the
>> JAX-WS MR. And also there'll be more places at which we can do improvements
>> to gain even better performance numbers.
>>
>> Thanks,
>> ~Isuru
>>
>> [1] https://issues.apache.org/jira/browse/AXIS2-4930
>>
>> On Tue, Nov 29, 2011 at 4:51 PM, Charith Wickramarachchi
>> <ch...@gmail.com> wrote:
>>>
>>> Hi Devs ,
>>>
>>> JAX-WS is becoming a popular key webservice development style. While
>>> supporting its own web-service development/deployment style i think Axis2
>>> should improve its JAX-WS implementation. I started to profile Axis2 JAX-WS
>>> implementation having performance improvement in mind and was able to
>>> isolate a major bottleneck in this flow. (See the attached profile image.)
>>> Here the bottleneck was AXOM creating a XMLOutput factory per each
>>> invocation due to a bug in JAXWSMessageReceiver. I submitted a fix for this
>>> in[1]. And we are currently in the process of discussing on a proper
>>> solution for that issue.
>>>
>>> With this patch i was able to get ~75% performance improvement on Simple
>>> Echo jax-ws services and ~60% improvement with services with complex types.
>>>
>>>
>>> With this observation i think if we spend some energy on improving Axis2
>>> JAX-WS performance we will be able to improve it a lot and
>>> if we can publish that improved numbers may be with comparisons that will
>>> add a great value to this project.
>>>
>>> So i though of starting a discussion around it to discuss on how to
>>> improve this and areas we must focus on.
>>>
>>>
>>> [1]https://issues.apache.org/jira/browse/AXIS2-5203
>>>
>>> thanks,
>>> Charith
>>>
>>> --
>>> Charith Dhanushka Wickramarachchi
>>> http://charithwiki.blogspot.com/
>>>
>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
>>> For additional commands, e-mail: java-dev-help@axis.apache.org
>>
>>
>>
>>
>> --
>> Technical Lead,
>> WSO2 Inc. http://wso2.org/
>> Blog : http://isurues.wordpress.com/
>
>
>
>
> --
> Sagara Gunathunga
>
> Blog - http://ssagara.blogspot.com
> Web - http://people.apache.org/~sagara/
> LinkedIn - http://www.linkedin.com/in/ssagara
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> For additional commands, e-mail: java-dev-help@axis.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
For additional commands, e-mail: java-dev-help@axis.apache.org
Re: Improving JAX-WS Performance
Posted by Sagara Gunathunga <sa...@gmail.com>.
I profiled a complex JAX-WS web service (refer the attachment) it seems 45%
of time taken for JAXB unmarshalling while another 28.9% of time taken for
message.getBodyBlockQNames() that eventually build the message through
AXIOM. I experimented hard coding some of the value calculation to avoid
this message building but what I realized is it's not a single place JAX-WS
build this message instead when i avoid message building in a one place it
build the message in a later stage.
I can't realize why we have to build the message instead of passing it to
JAXB Unmarshaller without building the message.
For me this seems like a extra cost , can any one explain the rational
behind this idea ? or can't we git rid of this message building.
Thanks !
On Wed, Nov 30, 2011 at 5:27 AM, Isuru Suriarachchi <is...@gmail.com>wrote:
> Hi Charith,
>
> +1 for putting some more effort on improving our JAX-WS support. I'll
> commit the proper fix to [1] ASAP and that will remove the bottleneck in
> the JAX-WS MR. And also there'll be more places at which we can do
> improvements to gain even better performance numbers.
>
> Thanks,
> ~Isuru
>
> [1] https://issues.apache.org/jira/browse/AXIS2-4930
>
> On Tue, Nov 29, 2011 at 4:51 PM, Charith Wickramarachchi <
> charith.dhanushka@gmail.com> wrote:
>
>> Hi Devs ,
>>
>> JAX-WS is becoming a popular key webservice development style. While
>> supporting its own web-service development/deployment style i think Axis2
>> should improve its JAX-WS implementation. I started to profile Axis2 JAX-WS
>> implementation having performance improvement in mind and was able to
>> isolate a major bottleneck in this flow. (See the attached profile image.)
>> Here the bottleneck was AXOM creating a XMLOutput factory per each
>> invocation due to a bug in JAXWSMessageReceiver. I submitted a fix for this
>> in[1]. And we are currently in the process of discussing on a proper
>> solution for that issue.
>>
>> With this patch i was able to get ~75% performance improvement on Simple
>> Echo jax-ws services and ~60% improvement with services with complex types.
>>
>>
>> With this observation i think if we spend some energy on improving Axis2
>> JAX-WS performance we will be able to improve it a lot and
>> if we can publish that improved numbers may be with comparisons that will
>> add a great value to this project.
>>
>> So i though of starting a discussion around it to discuss on how to
>> improve this and areas we must focus on.
>>
>>
>> [1]https://issues.apache.org/jira/browse/AXIS2-5203
>>
>> thanks,
>> Charith
>>
>> --
>> Charith Dhanushka Wickramarachchi
>> http://charithwiki.blogspot.com/
>>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
>> For additional commands, e-mail: java-dev-help@axis.apache.org
>>
>
>
>
> --
> Technical Lead,
> WSO2 Inc. http://wso2.org/
> Blog : http://isurues.wordpress.com/
>
--
Sagara Gunathunga
Blog - http://ssagara.blogspot.com
Web - http://people.apache.org/~sagara/
LinkedIn - http://www.linkedin.com/in/ssagara
Re: Improving JAX-WS Performance
Posted by Isuru Suriarachchi <is...@gmail.com>.
Hi Charith,
+1 for putting some more effort on improving our JAX-WS support. I'll
commit the proper fix to [1] ASAP and that will remove the bottleneck in
the JAX-WS MR. And also there'll be more places at which we can do
improvements to gain even better performance numbers.
Thanks,
~Isuru
[1] https://issues.apache.org/jira/browse/AXIS2-4930
On Tue, Nov 29, 2011 at 4:51 PM, Charith Wickramarachchi <
charith.dhanushka@gmail.com> wrote:
> Hi Devs ,
>
> JAX-WS is becoming a popular key webservice development style. While
> supporting its own web-service development/deployment style i think Axis2
> should improve its JAX-WS implementation. I started to profile Axis2 JAX-WS
> implementation having performance improvement in mind and was able to
> isolate a major bottleneck in this flow. (See the attached profile image.)
> Here the bottleneck was AXOM creating a XMLOutput factory per each
> invocation due to a bug in JAXWSMessageReceiver. I submitted a fix for this
> in[1]. And we are currently in the process of discussing on a proper
> solution for that issue.
>
> With this patch i was able to get ~75% performance improvement on Simple
> Echo jax-ws services and ~60% improvement with services with complex types.
>
>
> With this observation i think if we spend some energy on improving Axis2
> JAX-WS performance we will be able to improve it a lot and
> if we can publish that improved numbers may be with comparisons that will
> add a great value to this project.
>
> So i though of starting a discussion around it to discuss on how to
> improve this and areas we must focus on.
>
>
> [1]https://issues.apache.org/jira/browse/AXIS2-5203
>
> thanks,
> Charith
>
> --
> Charith Dhanushka Wickramarachchi
> http://charithwiki.blogspot.com/
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: java-dev-unsubscribe@axis.apache.org
> For additional commands, e-mail: java-dev-help@axis.apache.org
>
--
Technical Lead,
WSO2 Inc. http://wso2.org/
Blog : http://isurues.wordpress.com/