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 "Gudla, Natraj (GE Money, consultant)" <na...@ge.com> on 2007/12/11 09:48:43 UTC

Real time axis performance - Any case studies available?

Hi All,
 
Is Axis2 proven to handle real time heavy load of concurrent request(s)
and process them with quick responses. I am interested to check any case
studies or hard learnings from any one who might have had this
situation. We are trying to build an on-line payment application, which
requires customer payments to be authored in quick turn around seconds.
The web applicaiton will hit a new web service layer to be developed
which internally interacts with one or more third party web services. 
 
Before we kick off using Axis2 we need to understand the real time
performance. Any inputs will highly be appreciated.
 
Cheers
Natraj.

Re: Real time axis performance - Any case studies available?

Posted by robert lazarski <ro...@gmail.com>.
Not that I'm connected with this company in any way, but one option
that has helped our performance with ssl is a dedicated load balancer
that does hardware ssl acceleration - in my case the loadmaster
LMI-1500. Their docs describe this as:

"Putting a LoadMaster or a SSL-Master in front of these web servers
and terminating the
SSL session, thus sending the traffic as regular HTTP to the servers
relieves the servers
of the CPU-intensive cryptographic functions. Using a LoadMaster also allows the
ability to do cookie persistence, since the LoadMaster will see the
unencrypted traffic
and will be able to make persistence decisions based on cookies."

HTH,
Robert

On Dec 11, 2007 3:36 PM, Paul Fremantle <pz...@gmail.com> wrote:
> Its possible another JSSE implementation might be faster - you could try
> running on IBMs or BEAs JDKs. I guess you could find an SSL accelerator card
> that supports Java. One more thing worth trying is the async NIO HTTP
> transport that is in Axis2 1.3. It might help.
>
> Paul
>
>
>
> On Dec 11, 2007 5:13 PM, Joe S <jo...@gmail.com> wrote:
> > Paul,
> >
> > I tried your suggestion to reuse HTTPClient and it solved my problem!
> Thank you so much, you saved me! Https connection speed is now close to http
> after establishing the SSL session, but it's still slower than the .Net
> client or Jmeter. Is there anything else I can do to improve the performance
> using http or https?
> >
> > Thanks again.
> >
> > Joe
> >
> >
> >
> >
> >
> > On Dec 11, 2007 9:16 AM, Paul Fremantle < pzfreo@gmail.com> wrote:
> >
> > > Joe
> > >
> > > The article talks about JVM tuning and this article talks about TCP
> tuning - at least for Linux. http://wso2.org/library/1721
> > >
> > > I think the SSL problem may require you to re-use the same HTTPClient in
> your Axis2 client. I'm guessing that what's happening is that the SSL
> negotiation is happening with every single request.
> > >
> > > try
> stub._getServiceClient().getOptions().setProperty(HTTPConstants.REUSE_HTTP_CLIENT,
> "true");
> > >
> > > Paul
> > >
> > >
> > >
> > >
> > >
> > > On Dec 11, 2007 2:06 PM, Joe S <joe.smithian@gmail.com > wrote:
> > >
> > > > Hi Paul,
> > > >
> > > > Thanks for the good points. What tuning we should make to the TCP
> stack or JVM? Any reference for more information?
> > > >
> > > > What I found in my tests using AXIS2 is that using https in Java
> client  is unacceptably slow, every SOAP operation is 100 times slower than
> http. What can we do about it? Does that has anything to do with AXIS2
> configurations? or Sun's implementation of SSL?
> > > >
> > > > Thanks
> > > >
> > > > Joe
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > On Dec 11, 2007 4:21 AM, Paul Fremantle <pzfreo@gmail.com > wrote:
> > > >
> > > > > Natraj
> > > > >
> > > > > We did some significant benchmarking a while back that showed some
> pretty good results:
> > > > >
> > > > > http://wso2.org/library/588
> > > > >
> > > > > Under heavy load you will also need to tune your TCP stack and JVM
> to get good performance.
> > > > >
> > > > > Paul
> > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant)
> <natraj.gudla@ge.com > wrote:
> > > > >
> > > > > >
> > > > > >
> > > > > > Hi All,
> > > > > >
> > > > > > Is Axis2 proven to handle real time heavy load of concurrent
> request(s) and process them with quick responses. I am interested to check
> any case studies or hard learnings from any one who might have had this
> situation. We are trying to build an on-line payment application, which
> requires customer payments to be authored in quick turn around seconds. The
> web applicaiton will hit a new web service layer to be developed which
> internally interacts with one or more third party web services.
> > > > > >
> > > > > > Before we kick off using Axis2 we need to understand the real time
> performance. Any inputs will highly be appreciated.
> > > > > >
> > > > > > Cheers
> > > > > > Natraj.
> > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Paul Fremantle
> > > > > Co-Founder and VP of Technical Sales, WSO2
> > > > > OASIS WS-RX TC Co-chair
> > > > >
> > > > > blog: http://pzf.fremantle.org
> > > > > paul@wso2.com
> > > > >
> > > > > "Oxygenating the Web Service Platform", www.wso2.com
> > > >
> > > >
> > >
> > >
> > >
> > > --
> > > Paul Fremantle
> > > Co-Founder and VP of Technical Sales, WSO2
> > > OASIS WS-RX TC Co-chair
> > >
> > > blog: http://pzf.fremantle.org
> > > paul@wso2.com
> > >
> > > "Oxygenating the Web Service Platform", www.wso2.com
> >
> >
>
>
>
> --
> Paul Fremantle
> Co-Founder and VP of Technical Sales, WSO2
> OASIS WS-RX TC Co-chair
>
> blog: http://pzf.fremantle.org
> paul@wso2.com
>
> "Oxygenating the Web Service Platform", www.wso2.com

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


Re: Real time axis performance - Any case studies available?

Posted by Joe S <jo...@gmail.com>.
Thanks Paul, you mentioned a good point about async invocation which  was my
next question, whether asycn SOAP invocation can provide a better
performance results for time-consuming server operations.

Joe


On Dec 11, 2007 3:36 PM, Paul Fremantle <pz...@gmail.com> wrote:

> Its possible another JSSE implementation might be faster - you could try
> running on IBMs or BEAs JDKs. I guess you could find an SSL accelerator card
> that supports Java. One more thing worth trying is the async NIO HTTP
> transport that is in Axis2 1.3. It might help.
>
> Paul
>
>
> On Dec 11, 2007 5:13 PM, Joe S <jo...@gmail.com> wrote:
>
> > Paul,
> >
> > I tried your suggestion to reuse HTTPClient and it solved my problem!
> > Thank you so much, you saved me! Https connection speed is now close to http
> > after establishing the SSL session, but it's still slower than the .Net
> > client or Jmeter. Is there anything else I can do to improve the performance
> > using http or https?
> >
> > Thanks again.
> >
> > Joe
> >
> >
> > On Dec 11, 2007 9:16 AM, Paul Fremantle < pzfreo@gmail.com> wrote:
> >
> > > Joe
> > >
> > > The article talks about JVM tuning and this article talks about TCP
> > > tuning - at least for Linux. http://wso2.org/library/1721
> > >
> > > I think the SSL problem may require you to re-use the same HTTPClient
> > > in your Axis2 client. I'm guessing that what's happening is that the SSL
> > > negotiation is happening with every single request.
> > >
> > > try stub._getServiceClient().getOptions().setProperty(
> > > HTTPConstants.REUSE_HTTP_CLIENT, "true");
> > >
> > > Paul
> > >
> > >
> > > On Dec 11, 2007 2:06 PM, Joe S <joe.smithian@gmail.com > wrote:
> > >
> > > > Hi Paul,
> > > >
> > > > Thanks for the good points. What tuning we should make to the TCP
> > > > stack or JVM? Any reference for more information?
> > > >
> > > > What I found in my tests using AXIS2 is that using https in Java
> > > > client  is unacceptably slow, every SOAP operation is 100 times slower than
> > > > http. What can we do about it? Does that has anything to do with AXIS2
> > > > configurations? or Sun's implementation of SSL?
> > > >
> > > > Thanks
> > > >
> > > > Joe
> > > >
> > > >
> > > > On Dec 11, 2007 4:21 AM, Paul Fremantle <pzfreo@gmail.com > wrote:
> > > >
> > > > > Natraj
> > > > >
> > > > > We did some significant benchmarking a while back that showed some
> > > > > pretty good results:
> > > > >
> > > > > http://wso2.org/library/588
> > > > >
> > > > > Under heavy load you will also need to tune your TCP stack and JVM
> > > > > to get good performance.
> > > > >
> > > > > Paul
> > > > >
> > > > >
> > > > > On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant) <
> > > > > natraj.gudla@ge.com > wrote:
> > > > >
> > > > > >  Hi All,
> > > > > >
> > > > > > Is Axis2 proven to handle real time heavy load of concurrent
> > > > > > request(s) and process them with quick responses. I am interested to check
> > > > > > any case studies or hard learnings from any one who might have had this
> > > > > > situation. We are trying to build an on-line payment
> > > > > > application, which requires customer payments to be authored in quick turn
> > > > > > around seconds. The web applicaiton will hit a new web service layer to be
> > > > > > developed which internally interacts with one or more third party web
> > > > > > services.
> > > > > >
> > > > > > Before we kick off using Axis2 we need to understand the real
> > > > > > time performance. Any inputs will highly be appreciated.
> > > > > >
> > > > > > Cheers
> > > > > > Natraj.
> > > > > >
> > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Paul Fremantle
> > > > > Co-Founder and VP of Technical Sales, WSO2
> > > > > OASIS WS-RX TC Co-chair
> > > > >
> > > > > blog: http://pzf.fremantle.org
> > > > > paul@wso2.com
> > > > >
> > > > > "Oxygenating the Web Service Platform", www.wso2.com
> > > >
> > > >
> > > >
> > >
> > >
> > > --
> > > Paul Fremantle
> > > Co-Founder and VP of Technical Sales, WSO2
> > > OASIS WS-RX TC Co-chair
> > >
> > > blog: http://pzf.fremantle.org
> > > paul@wso2.com
> > >
> > > "Oxygenating the Web Service Platform", www.wso2.com
> > >
> >
> >
>
>
> --
> Paul Fremantle
> Co-Founder and VP of Technical Sales, WSO2
> OASIS WS-RX TC Co-chair
>
> blog: http://pzf.fremantle.org
> paul@wso2.com
>
> "Oxygenating the Web Service Platform", www.wso2.com
>

Re: Real time axis performance - Any case studies available?

Posted by Paul Fremantle <pz...@gmail.com>.
Its possible another JSSE implementation might be faster - you could try
running on IBMs or BEAs JDKs. I guess you could find an SSL accelerator card
that supports Java. One more thing worth trying is the async NIO HTTP
transport that is in Axis2 1.3. It might help.

Paul

On Dec 11, 2007 5:13 PM, Joe S <jo...@gmail.com> wrote:

> Paul,
>
> I tried your suggestion to reuse HTTPClient and it solved my problem!
> Thank you so much, you saved me! Https connection speed is now close to http
> after establishing the SSL session, but it's still slower than the .Net
> client or Jmeter. Is there anything else I can do to improve the performance
> using http or https?
>
> Thanks again.
>
> Joe
>
>
> On Dec 11, 2007 9:16 AM, Paul Fremantle <pz...@gmail.com> wrote:
>
> > Joe
> >
> > The article talks about JVM tuning and this article talks about TCP
> > tuning - at least for Linux. http://wso2.org/library/1721
> >
> > I think the SSL problem may require you to re-use the same HTTPClient in
> > your Axis2 client. I'm guessing that what's happening is that the SSL
> > negotiation is happening with every single request.
> >
> > try stub._getServiceClient().getOptions().setProperty(
> > HTTPConstants.REUSE_HTTP_CLIENT, "true");
> >
> > Paul
> >
> >
> > On Dec 11, 2007 2:06 PM, Joe S <joe.smithian@gmail.com > wrote:
> >
> > > Hi Paul,
> > >
> > > Thanks for the good points. What tuning we should make to the TCP
> > > stack or JVM? Any reference for more information?
> > >
> > > What I found in my tests using AXIS2 is that using https in Java
> > > client  is unacceptably slow, every SOAP operation is 100 times slower than
> > > http. What can we do about it? Does that has anything to do with AXIS2
> > > configurations? or Sun's implementation of SSL?
> > >
> > > Thanks
> > >
> > > Joe
> > >
> > >
> > > On Dec 11, 2007 4:21 AM, Paul Fremantle <pzfreo@gmail.com > wrote:
> > >
> > > > Natraj
> > > >
> > > > We did some significant benchmarking a while back that showed some
> > > > pretty good results:
> > > >
> > > > http://wso2.org/library/588
> > > >
> > > > Under heavy load you will also need to tune your TCP stack and JVM
> > > > to get good performance.
> > > >
> > > > Paul
> > > >
> > > >
> > > > On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant) <
> > > > natraj.gudla@ge.com > wrote:
> > > >
> > > > >  Hi All,
> > > > >
> > > > > Is Axis2 proven to handle real time heavy load of concurrent
> > > > > request(s) and process them with quick responses. I am interested to check
> > > > > any case studies or hard learnings from any one who might have had this
> > > > > situation. We are trying to build an on-line payment application,
> > > > > which requires customer payments to be authored in quick turn around
> > > > > seconds. The web applicaiton will hit a new web service layer to be
> > > > > developed which internally interacts with one or more third party web
> > > > > services.
> > > > >
> > > > > Before we kick off using Axis2 we need to understand the real time
> > > > > performance. Any inputs will highly be appreciated.
> > > > >
> > > > > Cheers
> > > > > Natraj.
> > > > >
> > > >
> > > >
> > > >
> > > > --
> > > > Paul Fremantle
> > > > Co-Founder and VP of Technical Sales, WSO2
> > > > OASIS WS-RX TC Co-chair
> > > >
> > > > blog: http://pzf.fremantle.org
> > > > paul@wso2.com
> > > >
> > > > "Oxygenating the Web Service Platform", www.wso2.com
> > >
> > >
> > >
> >
> >
> > --
> > Paul Fremantle
> > Co-Founder and VP of Technical Sales, WSO2
> > OASIS WS-RX TC Co-chair
> >
> > blog: http://pzf.fremantle.org
> > paul@wso2.com
> >
> > "Oxygenating the Web Service Platform", www.wso2.com
> >
>
>


-- 
Paul Fremantle
Co-Founder and VP of Technical Sales, WSO2
OASIS WS-RX TC Co-chair

blog: http://pzf.fremantle.org
paul@wso2.com

"Oxygenating the Web Service Platform", www.wso2.com

Re: Real time axis performance - Any case studies available?

Posted by Dennis Sosnoski <dm...@sosnoski.com>.
Hi Dave,

I haven't tried this, but judging from the info the first ("XML Suite")
is really designed around SAX/DOM style XML processing, which isn't the
main focus of web services work (the vast majority of web services use
data binding for converting data to and from XML). The second (the "SOA
Security Toolkit") looks a lot more interesting, especially if their
claims can be believed ("more than three times the performance compared
to competitive Open Source solutions").

It would surprise me if the "SOA Security Toolkit" really does perform
that well compared to the standard Rampart code. I've typically seen
about a 3-4x dropoff in performance (i.e., a sequence of message
exchanges takes 3-4x longer) when enabling encryption and signatures,
more like 2x dropoff in performance if only using encryption. Knowing
what actually has to be done to the message in order to generate a
signature I'd be surprised if the "SOA Security Toolkit" really does
perform more than three times better than the standard Rampart code -
that would be delivering about the same performance when using
WS-Security encryption as when using SSL without WS-Security, and give
full WS-Security signature and encryption with only a modest additional
overhead.

I'd like to try this out, but the license prohibits any discussion of
benchmarks - so even if I tried it I wouldn't be able to talk about it.
Doesn't exactly give me a lot of confidence that this will deliver the
claimed performance boost...

  - 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



Johnson, David E wrote:
> Have you looked at using Intel(r) XML Suite and Intel(r) SOA Security
> Toolkit (beta available with AXIS2 API) to enhance performance?
> www.intel.com/software/xml .
>
> Dave
>  
> -----Original Message-----
> From: Dennis Sosnoski [mailto:dms@sosnoski.com] 
> Sent: Tuesday, December 11, 2007 6:30 PM
> To: axis-user@ws.apache.org
> Subject: Re: Real time axis performance - Any case studies available?
>
> When I've run time comparisons using Axis2 with and without SSL I've
> found the SSL time is generally about 40-50% higher than the ordinary
> HTTP time. I haven't had to tweak anything on the client to obtain that
> speed, though I do use a self-signed certificate and a local trust store
> (passed to the client via a JVM arg -Djavax.net.ssl.truststore=...). I
> use Linux, and just confirmed this performance with Axis2 1.3 and the
> Sun 1.5 JVM.
>
> I don't think you're likely to get much better speed out of Axis2
> (unless you're using XMLBeans and can switch to ADB or JiBX binding -
> XMLBeans is significantly slower than either ADB or JiBX binding, which
> are about the same speed). If you really need faster web services, the
> fastest alternative I'm aware of is my own JibxSoap code
> (http://jibx.sourceforge.net/jibxsoap/index.html). This is somewhat out
> of date, but is in use by a number of organizations. It's generally at
> least twice as fast as Axis2, assuming you're passing significant
> amounts of data. I'm hoping to have a JiBX/WS replacement out soon which
> will offer equal or better performance and more features.
>
>   - Dennis
>
>   

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


RE: Real time axis performance - Any case studies available?

Posted by "Johnson, David E" <da...@intel.com>.
Have you looked at using Intel(r) XML Suite and Intel(r) SOA Security
Toolkit (beta available with AXIS2 API) to enhance performance?
www.intel.com/software/xml .

Dave
 
-----Original Message-----
From: Dennis Sosnoski [mailto:dms@sosnoski.com] 
Sent: Tuesday, December 11, 2007 6:30 PM
To: axis-user@ws.apache.org
Subject: Re: Real time axis performance - Any case studies available?

When I've run time comparisons using Axis2 with and without SSL I've
found the SSL time is generally about 40-50% higher than the ordinary
HTTP time. I haven't had to tweak anything on the client to obtain that
speed, though I do use a self-signed certificate and a local trust store
(passed to the client via a JVM arg -Djavax.net.ssl.truststore=...). I
use Linux, and just confirmed this performance with Axis2 1.3 and the
Sun 1.5 JVM.

I don't think you're likely to get much better speed out of Axis2
(unless you're using XMLBeans and can switch to ADB or JiBX binding -
XMLBeans is significantly slower than either ADB or JiBX binding, which
are about the same speed). If you really need faster web services, the
fastest alternative I'm aware of is my own JibxSoap code
(http://jibx.sourceforge.net/jibxsoap/index.html). This is somewhat out
of date, but is in use by a number of organizations. It's generally at
least twice as fast as Axis2, assuming you're passing significant
amounts of data. I'm hoping to have a JiBX/WS replacement out soon which
will offer equal or better performance and more features.

  - 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



Joe S wrote:
> Paul,
>
> I tried your suggestion to reuse HTTPClient and it solved my problem!
> Thank you so much, you saved me! Https connection speed is now close
> to http after establishing the SSL session, but it's still slower than
> the .Net client or Jmeter. Is there anything else I can do to improve
> the performance using http or https?
>
> Thanks again.
>
> Joe
>
> On Dec 11, 2007 9:16 AM, Paul Fremantle <pzfreo@gmail.com
> <ma...@gmail.com>> wrote:
>
>     Joe
>
>     The article talks about JVM tuning and this article talks about
>     TCP tuning - at least for Linux. http://wso2.org/library/1721
>
>     I think the SSL problem may require you to re-use the same
>     HTTPClient in your Axis2 client. I'm guessing that what's
>     happening is that the SSL negotiation is happening with every
>     single request.
>
>     try
>
stub._getServiceClient().getOptions().setProperty(HTTPConstants.REUSE_HT
TP_CLIENT,
>     "true");
>
>     Paul
>
>
>     On Dec 11, 2007 2:06 PM, Joe S <joe.smithian@gmail.com
>     <ma...@gmail.com>> wrote:
>
>         Hi Paul,
>
>         Thanks for the good points. What tuning we should make to the
>         TCP stack or JVM? Any reference for more information?
>
>         What I found in my tests using AXIS2 is that using https in
>         Java client  is unacceptably slow, every SOAP operation is 100
>         times slower than http. What can we do about it? Does that has
>         anything to do with AXIS2 configurations? or Sun's
>         implementation of SSL?
>
>         Thanks
>
>         Joe
>
>
>         On Dec 11, 2007 4:21 AM, Paul Fremantle <pzfreo@gmail.com
>         <ma...@gmail.com>> wrote:
>
>             Natraj
>
>             We did some significant benchmarking a while back that
>             showed some pretty good results:
>
>             http://wso2.org/library/588
>
>             Under heavy load you will also need to tune your TCP stack
>             and JVM to get good performance.
>
>             Paul
>
>
>             On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money,
>             consultant) <natraj.gudla@ge.com
>             <ma...@ge.com> > wrote:
>
>                 Hi All,
>                  
>                 Is Axis2 proven to handle real time heavy load of
>                 concurrent request(s) and process them with quick
>                 responses. I am interested to check any case studies
>                 or hard learnings from any one who might have had this
>                 situation. We are trying to build an on-line payment
>                 application, which requires customer payments to be
>                 authored in quick turn around seconds. The web
>                 applicaiton will hit a new web service layer to be
>                 developed which internally interacts with one or more
>                 third party web services.
>                  
>                 Before we kick off using Axis2 we need to understand
>                 the real time performance. Any inputs will highly be
>                 appreciated.
>                  
>                 Cheers
>                 Natraj.
>
>
>
>
>             -- 
>             Paul Fremantle
>             Co-Founder and VP of Technical Sales, WSO2
>             OASIS WS-RX TC Co-chair
>
>             blog: http://pzf.fremantle.org <http://pzf.fremantle.org>
>             paul@wso2.com <ma...@wso2.com>
>
>             "Oxygenating the Web Service Platform", www.wso2.com
>             <http://www.wso2.com> 
>
>
>
>
>
>     -- 
>     Paul Fremantle
>     Co-Founder and VP of Technical Sales, WSO2
>     OASIS WS-RX TC Co-chair
>
>     blog: http://pzf.fremantle.org <http://pzf.fremantle.org>
>     paul@wso2.com <ma...@wso2.com>
>
>     "Oxygenating the Web Service Platform", www.wso2.com
>     <http://www.wso2.com>
>
>

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

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


Re: Real time axis performance - Any case studies available?

Posted by Dennis Sosnoski <dm...@sosnoski.com>.
When I've run time comparisons using Axis2 with and without SSL I've
found the SSL time is generally about 40-50% higher than the ordinary
HTTP time. I haven't had to tweak anything on the client to obtain that
speed, though I do use a self-signed certificate and a local trust store
(passed to the client via a JVM arg -Djavax.net.ssl.truststore=...). I
use Linux, and just confirmed this performance with Axis2 1.3 and the
Sun 1.5 JVM.

I don't think you're likely to get much better speed out of Axis2
(unless you're using XMLBeans and can switch to ADB or JiBX binding -
XMLBeans is significantly slower than either ADB or JiBX binding, which
are about the same speed). If you really need faster web services, the
fastest alternative I'm aware of is my own JibxSoap code
(http://jibx.sourceforge.net/jibxsoap/index.html). This is somewhat out
of date, but is in use by a number of organizations. It's generally at
least twice as fast as Axis2, assuming you're passing significant
amounts of data. I'm hoping to have a JiBX/WS replacement out soon which
will offer equal or better performance and more features.

  - 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



Joe S wrote:
> Paul,
>
> I tried your suggestion to reuse HTTPClient and it solved my problem!
> Thank you so much, you saved me! Https connection speed is now close
> to http after establishing the SSL session, but it's still slower than
> the .Net client or Jmeter. Is there anything else I can do to improve
> the performance using http or https?
>
> Thanks again.
>
> Joe
>
> On Dec 11, 2007 9:16 AM, Paul Fremantle <pzfreo@gmail.com
> <ma...@gmail.com>> wrote:
>
>     Joe
>
>     The article talks about JVM tuning and this article talks about
>     TCP tuning - at least for Linux. http://wso2.org/library/1721
>
>     I think the SSL problem may require you to re-use the same
>     HTTPClient in your Axis2 client. I'm guessing that what's
>     happening is that the SSL negotiation is happening with every
>     single request.
>
>     try
>     stub._getServiceClient().getOptions().setProperty(HTTPConstants.REUSE_HTTP_CLIENT,
>     "true");
>
>     Paul
>
>
>     On Dec 11, 2007 2:06 PM, Joe S <joe.smithian@gmail.com
>     <ma...@gmail.com>> wrote:
>
>         Hi Paul,
>
>         Thanks for the good points. What tuning we should make to the
>         TCP stack or JVM? Any reference for more information?
>
>         What I found in my tests using AXIS2 is that using https in
>         Java client  is unacceptably slow, every SOAP operation is 100
>         times slower than http. What can we do about it? Does that has
>         anything to do with AXIS2 configurations? or Sun's
>         implementation of SSL?
>
>         Thanks
>
>         Joe
>
>
>         On Dec 11, 2007 4:21 AM, Paul Fremantle <pzfreo@gmail.com
>         <ma...@gmail.com>> wrote:
>
>             Natraj
>
>             We did some significant benchmarking a while back that
>             showed some pretty good results:
>
>             http://wso2.org/library/588
>
>             Under heavy load you will also need to tune your TCP stack
>             and JVM to get good performance.
>
>             Paul
>
>
>             On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money,
>             consultant) <natraj.gudla@ge.com
>             <ma...@ge.com> > wrote:
>
>                 Hi All,
>                  
>                 Is Axis2 proven to handle real time heavy load of
>                 concurrent request(s) and process them with quick
>                 responses. I am interested to check any case studies
>                 or hard learnings from any one who might have had this
>                 situation. We are trying to build an on-line payment
>                 application, which requires customer payments to be
>                 authored in quick turn around seconds. The web
>                 applicaiton will hit a new web service layer to be
>                 developed which internally interacts with one or more
>                 third party web services.
>                  
>                 Before we kick off using Axis2 we need to understand
>                 the real time performance. Any inputs will highly be
>                 appreciated.
>                  
>                 Cheers
>                 Natraj.
>
>
>
>
>             -- 
>             Paul Fremantle
>             Co-Founder and VP of Technical Sales, WSO2
>             OASIS WS-RX TC Co-chair
>
>             blog: http://pzf.fremantle.org <http://pzf.fremantle.org>
>             paul@wso2.com <ma...@wso2.com>
>
>             "Oxygenating the Web Service Platform", www.wso2.com
>             <http://www.wso2.com> 
>
>
>
>
>
>     -- 
>     Paul Fremantle
>     Co-Founder and VP of Technical Sales, WSO2
>     OASIS WS-RX TC Co-chair
>
>     blog: http://pzf.fremantle.org <http://pzf.fremantle.org>
>     paul@wso2.com <ma...@wso2.com>
>
>     "Oxygenating the Web Service Platform", www.wso2.com
>     <http://www.wso2.com>
>
>

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


Re: Real time axis performance - Any case studies available?

Posted by Joe S <jo...@gmail.com>.
Paul,

I tried your suggestion to reuse HTTPClient and it solved my problem! Thank
you so much, you saved me! Https connection speed is now close to http after
establishing the SSL session, but it's still slower than the .Net client or
Jmeter. Is there anything else I can do to improve the performance using
http or https?

Thanks again.

Joe

On Dec 11, 2007 9:16 AM, Paul Fremantle <pz...@gmail.com> wrote:

> Joe
>
> The article talks about JVM tuning and this article talks about TCP tuning
> - at least for Linux. http://wso2.org/library/1721
>
> I think the SSL problem may require you to re-use the same HTTPClient in
> your Axis2 client. I'm guessing that what's happening is that the SSL
> negotiation is happening with every single request.
>
> try stub._getServiceClient().getOptions().setProperty(
> HTTPConstants.REUSE_HTTP_CLIENT, "true");
>
> Paul
>
>
> On Dec 11, 2007 2:06 PM, Joe S <jo...@gmail.com> wrote:
>
> > Hi Paul,
> >
> > Thanks for the good points. What tuning we should make to the TCP stack
> > or JVM? Any reference for more information?
> >
> > What I found in my tests using AXIS2 is that using https in Java client
> > is unacceptably slow, every SOAP operation is 100 times slower than http.
> > What can we do about it? Does that has anything to do with AXIS2
> > configurations? or Sun's implementation of SSL?
> >
> > Thanks
> >
> > Joe
> >
> >
> > On Dec 11, 2007 4:21 AM, Paul Fremantle <pzfreo@gmail.com > wrote:
> >
> > > Natraj
> > >
> > > We did some significant benchmarking a while back that showed some
> > > pretty good results:
> > >
> > > http://wso2.org/library/588
> > >
> > > Under heavy load you will also need to tune your TCP stack and JVM to
> > > get good performance.
> > >
> > > Paul
> > >
> > >
> > > On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant) <
> > > natraj.gudla@ge.com > wrote:
> > >
> > > >  Hi All,
> > > >
> > > > Is Axis2 proven to handle real time heavy load of concurrent
> > > > request(s) and process them with quick responses. I am interested to check
> > > > any case studies or hard learnings from any one who might have had this
> > > > situation. We are trying to build an on-line payment application,
> > > > which requires customer payments to be authored in quick turn around
> > > > seconds. The web applicaiton will hit a new web service layer to be
> > > > developed which internally interacts with one or more third party web
> > > > services.
> > > >
> > > > Before we kick off using Axis2 we need to understand the real time
> > > > performance. Any inputs will highly be appreciated.
> > > >
> > > > Cheers
> > > > Natraj.
> > > >
> > >
> > >
> > >
> > > --
> > > Paul Fremantle
> > > Co-Founder and VP of Technical Sales, WSO2
> > > OASIS WS-RX TC Co-chair
> > >
> > > blog: http://pzf.fremantle.org
> > > paul@wso2.com
> > >
> > > "Oxygenating the Web Service Platform", www.wso2.com
> >
> >
> >
>
>
> --
> Paul Fremantle
> Co-Founder and VP of Technical Sales, WSO2
> OASIS WS-RX TC Co-chair
>
> blog: http://pzf.fremantle.org
> paul@wso2.com
>
> "Oxygenating the Web Service Platform", www.wso2.com
>

Re: Real time axis performance - Any case studies available?

Posted by Paul Fremantle <pz...@gmail.com>.
Joe

The article talks about JVM tuning and this article talks about TCP tuning -
at least for Linux. http://wso2.org/library/1721

I think the SSL problem may require you to re-use the same HTTPClient in
your Axis2 client. I'm guessing that what's happening is that the SSL
negotiation is happening with every single request.

try stub._getServiceClient().getOptions().setProperty(
HTTPConstants.REUSE_HTTP_CLIENT, "true");

Paul

On Dec 11, 2007 2:06 PM, Joe S <jo...@gmail.com> wrote:

> Hi Paul,
>
> Thanks for the good points. What tuning we should make to the TCP stack or
> JVM? Any reference for more information?
>
> What I found in my tests using AXIS2 is that using https in Java client
> is unacceptably slow, every SOAP operation is 100 times slower than http.
> What can we do about it? Does that has anything to do with AXIS2
> configurations? or Sun's implementation of SSL?
>
> Thanks
>
> Joe
>
>
> On Dec 11, 2007 4:21 AM, Paul Fremantle <pz...@gmail.com> wrote:
>
> > Natraj
> >
> > We did some significant benchmarking a while back that showed some
> > pretty good results:
> >
> > http://wso2.org/library/588
> >
> > Under heavy load you will also need to tune your TCP stack and JVM to
> > get good performance.
> >
> > Paul
> >
> >
> > On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant) <
> > natraj.gudla@ge.com > wrote:
> >
> > >  Hi All,
> > >
> > > Is Axis2 proven to handle real time heavy load of concurrent
> > > request(s) and process them with quick responses. I am interested to check
> > > any case studies or hard learnings from any one who might have had this
> > > situation. We are trying to build an on-line payment application,
> > > which requires customer payments to be authored in quick turn around
> > > seconds. The web applicaiton will hit a new web service layer to be
> > > developed which internally interacts with one or more third party web
> > > services.
> > >
> > > Before we kick off using Axis2 we need to understand the real time
> > > performance. Any inputs will highly be appreciated.
> > >
> > > Cheers
> > > Natraj.
> > >
> >
> >
> >
> > --
> > Paul Fremantle
> > Co-Founder and VP of Technical Sales, WSO2
> > OASIS WS-RX TC Co-chair
> >
> > blog: http://pzf.fremantle.org
> > paul@wso2.com
> >
> > "Oxygenating the Web Service Platform", www.wso2.com
>
>
>


-- 
Paul Fremantle
Co-Founder and VP of Technical Sales, WSO2
OASIS WS-RX TC Co-chair

blog: http://pzf.fremantle.org
paul@wso2.com

"Oxygenating the Web Service Platform", www.wso2.com

Re: Real time axis performance - Any case studies available?

Posted by Joe S <jo...@gmail.com>.
Hi Paul,

Thanks for the good points. What tuning we should make to the TCP stack or
JVM? Any reference for more information?

What I found in my tests using AXIS2 is that using https in Java client  is
unacceptably slow, every SOAP operation is 100 times slower than http. What
can we do about it? Does that has anything to do with AXIS2 configurations?
or Sun's implementation of SSL?

Thanks

Joe

On Dec 11, 2007 4:21 AM, Paul Fremantle <pz...@gmail.com> wrote:

> Natraj
>
> We did some significant benchmarking a while back that showed some pretty
> good results:
>
> http://wso2.org/library/588
>
> Under heavy load you will also need to tune your TCP stack and JVM to get
> good performance.
>
> Paul
>
>
> On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant) <
> natraj.gudla@ge.com> wrote:
>
> >  Hi All,
> >
> > Is Axis2 proven to handle real time heavy load of concurrent request(s)
> > and process them with quick responses. I am interested to check any case
> > studies or hard learnings from any one who might have had this situation. We
> > are trying to build an on-line payment application, which requires customer
> > payments to be authored in quick turn around seconds. The web applicaiton
> > will hit a new web service layer to be developed which internally interacts
> > with one or more third party web services.
> >
> > Before we kick off using Axis2 we need to understand the real time
> > performance. Any inputs will highly be appreciated.
> >
> > Cheers
> > Natraj.
> >
>
>
>
> --
> Paul Fremantle
> Co-Founder and VP of Technical Sales, WSO2
> OASIS WS-RX TC Co-chair
>
> blog: http://pzf.fremantle.org
> paul@wso2.com
>
> "Oxygenating the Web Service Platform", www.wso2.com

RE: Real time axis performance - Any case studies available?

Posted by "Gudla, Natraj (GE Money, consultant)" <na...@ge.com>.
Thanks Paul.

________________________________

From: Paul Fremantle [mailto:pzfreo@gmail.com] 
Sent: 11 December 2007 09:22
To: axis-user@ws.apache.org
Subject: Re: Real time axis performance - Any case studies available?


Natraj

We did some significant benchmarking a while back that showed some
pretty good results:

http://wso2.org/library/588

Under heavy load you will also need to tune your TCP stack and JVM to
get good performance. 

Paul


On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant)
<na...@ge.com> wrote:


	Hi All,
	 
	Is Axis2 proven to handle real time heavy load of concurrent
request(s) and process them with quick responses. I am interested to
check any case studies or hard learnings from any one who might have had
this situation. We are trying to build an on-line payment application,
which requires customer payments to be authored in quick turn around
seconds. The web applicaiton will hit a new web service layer to be
developed which internally interacts with one or more third party web
services. 
	 
	Before we kick off using Axis2 we need to understand the real
time performance. Any inputs will highly be appreciated.
	 
	Cheers
	Natraj.




-- 
Paul Fremantle
Co-Founder and VP of Technical Sales, WSO2
OASIS WS-RX TC Co-chair

blog: http://pzf.fremantle.org
paul@wso2.com

"Oxygenating the Web Service Platform", www.wso2.com 

Re: Real time axis performance - Any case studies available?

Posted by Paul Fremantle <pz...@gmail.com>.
Natraj

We did some significant benchmarking a while back that showed some pretty
good results:

http://wso2.org/library/588

Under heavy load you will also need to tune your TCP stack and JVM to get
good performance.

Paul

On Dec 11, 2007 8:48 AM, Gudla, Natraj (GE Money, consultant) <
natraj.gudla@ge.com> wrote:

>  Hi All,
>
> Is Axis2 proven to handle real time heavy load of concurrent request(s)
> and process them with quick responses. I am interested to check any case
> studies or hard learnings from any one who might have had this situation. We
> are trying to build an on-line payment application, which requires customer
> payments to be authored in quick turn around seconds. The web applicaiton
> will hit a new web service layer to be developed which internally interacts
> with one or more third party web services.
>
> Before we kick off using Axis2 we need to understand the real time
> performance. Any inputs will highly be appreciated.
>
> Cheers
> Natraj.
>



-- 
Paul Fremantle
Co-Founder and VP of Technical Sales, WSO2
OASIS WS-RX TC Co-chair

blog: http://pzf.fremantle.org
paul@wso2.com

"Oxygenating the Web Service Platform", www.wso2.com

RE: Real time axis performance - Any case studies available?

Posted by "Gudla, Natraj (GE Money, consultant)" <na...@ge.com>.
Thanks Dennis, that was a good information.  

-----Original Message-----
From: Dennis Sosnoski [mailto:dms@sosnoski.com] 
Sent: 11 December 2007 09:15
To: axis-user@ws.apache.org
Subject: Re: Real time axis performance - Any case studies available?

Hi Natraj,

Response under heavy load is going to depend on a number of issues,
including the servlet engine used for deployment. Since you're talking
about payments, security may also be an issue - WS-Security processing
overhead will generally cut your performance by a factor of 2-4x
(depending on the size of your data, and on whether you're using both
encryption and signatures or only encryption). From what I've seen,
though, Axis2 performance is pretty much on a par with other comparable
frameworks (JAX-WS, XFire).

The best way to check performance for your specific requirements is to
mock-up the service and try it out under load. If you return a somewhat
realistic response to somewhat realistic requests you should get a very
good idea of the Axis2 overhead. Then you just need to judge whether
there's enough processor time available for running your actual
implementation code. As far as I'm concerned, doing this kind of test is
a best practice for any applications with specific performance
requirements.

  - 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



Gudla, Natraj (GE Money, consultant) wrote:
> Hi All,
>  
> Is Axis2 proven to handle real time heavy load of concurrent
> request(s) and process them with quick responses. I am interested to
> check any case studies or hard learnings from any one who might have
> had this situation. We are trying to build an on-line payment
> application, which requires customer payments to be authored in quick
> turn around seconds. The web applicaiton will hit a new web service
> layer to be developed which internally interacts with one or more
> third party web services.
>  
> Before we kick off using Axis2 we need to understand the real time
> performance. Any inputs will highly be appreciated.
>  
> Cheers
> Natraj.

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


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


Re: Real time axis performance - Any case studies available?

Posted by Dennis Sosnoski <dm...@sosnoski.com>.
Hi Natraj,

Response under heavy load is going to depend on a number of issues,
including the servlet engine used for deployment. Since you're talking
about payments, security may also be an issue - WS-Security processing
overhead will generally cut your performance by a factor of 2-4x
(depending on the size of your data, and on whether you're using both
encryption and signatures or only encryption). From what I've seen,
though, Axis2 performance is pretty much on a par with other comparable
frameworks (JAX-WS, XFire).

The best way to check performance for your specific requirements is to
mock-up the service and try it out under load. If you return a somewhat
realistic response to somewhat realistic requests you should get a very
good idea of the Axis2 overhead. Then you just need to judge whether
there's enough processor time available for running your actual
implementation code. As far as I'm concerned, doing this kind of test is
a best practice for any applications with specific performance requirements.

  - 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



Gudla, Natraj (GE Money, consultant) wrote:
> Hi All,
>  
> Is Axis2 proven to handle real time heavy load of concurrent
> request(s) and process them with quick responses. I am interested to
> check any case studies or hard learnings from any one who might have
> had this situation. We are trying to build an on-line payment
> application, which requires customer payments to be authored in quick
> turn around seconds. The web applicaiton will hit a new web service
> layer to be developed which internally interacts with one or more
> third party web services.
>  
> Before we kick off using Axis2 we need to understand the real time
> performance. Any inputs will highly be appreciated.
>  
> Cheers
> Natraj.

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