You are viewing a plain text version of this content. The canonical link for it is here.
Posted to java-user@axis.apache.org by Sardar Hussain <sa...@yahoo.com.INVALID> on 2014/11/29 21:07:51 UTC

Axis2 Performance Issue

Hi,
I have a strange issue with performance evaluation using Axis2 middleware. I have two scenarios of calling web services through a client. One in centralised fashion and the other in decentralised fashion. Centralised scenario 1 goes like this: I have a client that dynamically invokes 5 different services in a workflow fashion. I have engaged a custom module as inflow handler to the 5 different services. I calculate the performance evaluation of calling these 5 services with the engagement of my module as well as without engaging the module. When I call these services through the same client as a workflow, without my custom module then it takes 1912 milli seconds. However, when I engage my custom module to the services it takes 2070- milli seconds. This shows the performance overhead of my custom module equivalent to 160 milli seconds.
The second secnario 2 of decentralised invocation of the same services goes like this:A client invokes the first service of the 5 different services. Then the first service invokes the second service and similarly the sequence goes to the invocation of service 5 through the service 4. These services are invoked asynchronously (withoutout blocking) as a workflow. These services perform their own functionalities like in the earlier scenario, however, they are playing the role of a client for the invocation of the next service.Now the performance I get with the engagement and without engagement of my custom module goes like this:Without engageing the module it takes ~ 1450 milli second, which is less than 1850 as was the case in scenario 1, because they are invoked in decentralised fashion and hence no centralised bottleneck.But when I invoke this decentralised invocation with the engagement of the same module, it takes more than 2094 milli second. which shows that the module I have engaged is taking more than 600 milli second. I am wondering this module is the same as was engaged in scenario 1 and it took 160 milli second.  However, here it takes much more (~ 600 milli second) compared to the centralised one. Can one kindly guide me through where and what the reason might be?
regards,
Sardar Hussain

Re: Axis2 Performance Issue

Posted by Navinderjit Kahlon <na...@gmail.com>.
hi...

i m also working on same topic to call web service dynamically but not able
to achieve it so far...

could u plz help me how u have performed dynamic invocation of web
service... the complete scenario and its entire development....

On Sun, Nov 30, 2014 at 1:37 AM, Sardar Hussain <
salarzi_qau@yahoo.com.invalid> wrote:

> Hi,
>
> I have a strange issue with performance evaluation using Axis2 middleware.
>  I have two scenarios of calling web services through a client. One in
> centralised fashion and the other in decentralised fashion. Centralised
> scenario 1 goes like this:
>  I have a client that dynamically invokes 5 different services in a
> workflow fashion. I have engaged a custom module as inflow handler to the 5
> different services. I calculate the performance evaluation of calling these
> 5 services with the engagement of my module as well as without engaging the
> module.
> When I call these services through the same client as a workflow, without
> my custom module then it takes 1912 milli seconds. However, when I engage
> my custom module to the services it takes 2070- milli seconds. This shows
> the performance overhead of my custom module equivalent to 160 milli
> seconds.
>
> The second secnario 2 of decentralised invocation of the same services
> goes like this:
> A client invokes the first service of the 5 different services. Then the
> first service invokes the second service and similarly the sequence goes to
> the invocation of service 5 through the service 4. These services are
> invoked asynchronously (withoutout blocking) as a workflow. These services
> perform their own functionalities like in the earlier scenario, however,
> they are playing the role of a client for the invocation of the next
> service.
> Now the performance I get with the engagement and without engagement of my
> custom module goes like this:
> Without engageing the module it takes ~ 1450 milli second, which is less
> than 1850 as was the case in scenario 1, because they are invoked in
> decentralised fashion and hence no centralised bottleneck.
> But when I invoke this decentralised invocation with the engagement of the
> same module, it takes more than 2094 milli second. which shows that the
> module I have engaged is taking more than 600 milli second. I am wondering
> this module is the same as was engaged in scenario 1 and it took 160 milli
> second.  However, here it takes much more (~ 600 milli second) compared to
> the centralised one. Can one kindly guide me through where and what the
> reason might be?
>
> regards,
>
> Sardar Hussain
>
>


-- 
With thanks and regards

Navinderjit Kaur