You are viewing a plain text version of this content. The canonical link for it is here.
Posted to soap-user@ws.apache.org by sachin chaudhari <ch...@yahoo.com> on 2001/06/21 04:03:26 UTC
web services performance
Hi,
I did a simple test to see how much performance
penalty one will have to pay for Openness of SOAP.
RMI was 6 times faster than web services. I used a
very simple function call to test this. (I am not
claiming that I am right, but this is what I got!!!)
I am sure other architects are also facing the
challenge of deciding what level of services one needs
to expose as web services. If we decide to expose
fundamental services then performace of the
applications will be very slow.
I am very interested in knowing what you think about
the future of web services performance? Any benchmark
data available to compare web services & other
protocals like CORBA or RMI or servlets.
Any comments will be highly appreciated.
S
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
thanks!
-----Original Message-----
From: Rob Jellinghaus [mailto:robj@unrealities.com]
Sent: Wednesday, June 20, 2001 11:38 PM
To: soap-user@xml.apache.org; soap-user@xml.apache.org
Subject: RE: web services performance
I don't mean to speak for the Axis team, and don't hold me / us to this,
but there's been some discussion recently about doing an 0.5 release in a
couple of weeks. Axis is definitely making progress :-) I'll forward this
question to axis-dev.
Cheers!
Rob
At 11:13 PM 6/20/2001 -0500, graham glass wrote:
>hi sanjiva,
>
>any idea when Axis will be available? i'm writing a book
>on web services for prentice hall and already have chapters
>on J2EE/SOAP and .NET/SOAP. it would be great to include
>a chapter on AXIS/SOAP!
>
>cheers,
>graham
>
>-----Original Message-----
>From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
>Sent: Wednesday, June 20, 2001 11:02 PM
>To: soap-user@xml.apache.org
>Subject: Re: web services performance
>
>
>> oh, so 5x wouldn't be good enough, huh? ;-)
>>
>> - graham
>
>:-) It certainly would be .. however, you'd have to put the
>fastest possible JAXP parser with Apache SOAP and then try
>it .. Xerces is solid and all, but ain't the fastest by
>any means.
>
>Apache SOAP has never had any optimization effort put into
>it. I am currently making some mods that'll improve things
>a bit (someone found an obvious bottleneck with the way
>SOAPMappingRegistry's were being created) but other than
>that we never put any effort into making it fast. Axis is
>charged with that mission .. Apache SOAP is a robust
>implementation (given the pretty large number of downloads
>its pretty impressive to not have any major bugs so far)
>and not one that concentrated on performance.
>
>Sanjiva.
>
>
>
>
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
thanks!
-----Original Message-----
From: Rob Jellinghaus [mailto:robj@unrealities.com]
Sent: Wednesday, June 20, 2001 11:38 PM
To: soap-user@xml.apache.org; soap-user@xml.apache.org
Subject: RE: web services performance
I don't mean to speak for the Axis team, and don't hold me / us to this,
but there's been some discussion recently about doing an 0.5 release in a
couple of weeks. Axis is definitely making progress :-) I'll forward this
question to axis-dev.
Cheers!
Rob
At 11:13 PM 6/20/2001 -0500, graham glass wrote:
>hi sanjiva,
>
>any idea when Axis will be available? i'm writing a book
>on web services for prentice hall and already have chapters
>on J2EE/SOAP and .NET/SOAP. it would be great to include
>a chapter on AXIS/SOAP!
>
>cheers,
>graham
>
>-----Original Message-----
>From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
>Sent: Wednesday, June 20, 2001 11:02 PM
>To: soap-user@xml.apache.org
>Subject: Re: web services performance
>
>
>> oh, so 5x wouldn't be good enough, huh? ;-)
>>
>> - graham
>
>:-) It certainly would be .. however, you'd have to put the
>fastest possible JAXP parser with Apache SOAP and then try
>it .. Xerces is solid and all, but ain't the fastest by
>any means.
>
>Apache SOAP has never had any optimization effort put into
>it. I am currently making some mods that'll improve things
>a bit (someone found an obvious bottleneck with the way
>SOAPMappingRegistry's were being created) but other than
>that we never put any effort into making it fast. Axis is
>charged with that mission .. Apache SOAP is a robust
>implementation (given the pretty large number of downloads
>its pretty impressive to not have any major bugs so far)
>and not one that concentrated on performance.
>
>Sanjiva.
>
>
>
>
RE: web services performance
Posted by Rob Jellinghaus <ro...@unrealities.com>.
I don't mean to speak for the Axis team, and don't hold me / us to this,
but there's been some discussion recently about doing an 0.5 release in a
couple of weeks. Axis is definitely making progress :-) I'll forward this
question to axis-dev.
Cheers!
Rob
At 11:13 PM 6/20/2001 -0500, graham glass wrote:
>hi sanjiva,
>
>any idea when Axis will be available? i'm writing a book
>on web services for prentice hall and already have chapters
>on J2EE/SOAP and .NET/SOAP. it would be great to include
>a chapter on AXIS/SOAP!
>
>cheers,
>graham
>
>-----Original Message-----
>From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
>Sent: Wednesday, June 20, 2001 11:02 PM
>To: soap-user@xml.apache.org
>Subject: Re: web services performance
>
>
>> oh, so 5x wouldn't be good enough, huh? ;-)
>>
>> - graham
>
>:-) It certainly would be .. however, you'd have to put the
>fastest possible JAXP parser with Apache SOAP and then try
>it .. Xerces is solid and all, but ain't the fastest by
>any means.
>
>Apache SOAP has never had any optimization effort put into
>it. I am currently making some mods that'll improve things
>a bit (someone found an obvious bottleneck with the way
>SOAPMappingRegistry's were being created) but other than
>that we never put any effort into making it fast. Axis is
>charged with that mission .. Apache SOAP is a robust
>implementation (given the pretty large number of downloads
>its pretty impressive to not have any major bugs so far)
>and not one that concentrated on performance.
>
>Sanjiva.
>
>
>
>
RE: web services performance
Posted by Rob Jellinghaus <ro...@unrealities.com>.
I don't mean to speak for the Axis team, and don't hold me / us to this,
but there's been some discussion recently about doing an 0.5 release in a
couple of weeks. Axis is definitely making progress :-) I'll forward this
question to axis-dev.
Cheers!
Rob
At 11:13 PM 6/20/2001 -0500, graham glass wrote:
>hi sanjiva,
>
>any idea when Axis will be available? i'm writing a book
>on web services for prentice hall and already have chapters
>on J2EE/SOAP and .NET/SOAP. it would be great to include
>a chapter on AXIS/SOAP!
>
>cheers,
>graham
>
>-----Original Message-----
>From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
>Sent: Wednesday, June 20, 2001 11:02 PM
>To: soap-user@xml.apache.org
>Subject: Re: web services performance
>
>
>> oh, so 5x wouldn't be good enough, huh? ;-)
>>
>> - graham
>
>:-) It certainly would be .. however, you'd have to put the
>fastest possible JAXP parser with Apache SOAP and then try
>it .. Xerces is solid and all, but ain't the fastest by
>any means.
>
>Apache SOAP has never had any optimization effort put into
>it. I am currently making some mods that'll improve things
>a bit (someone found an obvious bottleneck with the way
>SOAPMappingRegistry's were being created) but other than
>that we never put any effort into making it fast. Axis is
>charged with that mission .. Apache SOAP is a robust
>implementation (given the pretty large number of downloads
>its pretty impressive to not have any major bugs so far)
>and not one that concentrated on performance.
>
>Sanjiva.
>
>
>
>
Re: web services performance
Posted by Sanjiva Weerawarana <sa...@watson.ibm.com>.
I certainly expect Axis to outperform to Apache SOAP. However,
comparing Axis and Apache SOAP right now is not useful (IMO)
until Axis supports SOAP Attachments too.
I believe GLUE does support SOAP Attachments.
Sanjiva.
----- Original Message -----
From: "Glen Daniels" <gd...@macromedia.com>
To: <so...@xml.apache.org>
Sent: Thursday, June 21, 2001 1:37 AM
Subject: Re: web services performance
> Just as an FYI, Axis on two of my machines benchmarked at about 4X and
about
> 8X the performance of ApacheSOAP 2.2 a few weeks ago (the 8X was certainly
> due to the fact that something in Axis (likely the parser) was able to
> utilize the dual processors on my work box to great effect). Still don't
> get no 500 msgs/second, though.... :) That said, Axis as it stands hasn't
> been seriously combed for optimizations either.
>
> --Glen
>
> ----- Original Message -----
> From: "Sanjiva Weerawarana" <sa...@watson.ibm.com>
> To: <so...@xml.apache.org>
> Sent: Thursday, June 21, 2001 12:01 AM
> Subject: Re: web services performance
>
>
> > > oh, so 5x wouldn't be good enough, huh? ;-)
> > >
> > > - graham
> >
> > :-) It certainly would be .. however, you'd have to put the
> > fastest possible JAXP parser with Apache SOAP and then try
> > it .. Xerces is solid and all, but ain't the fastest by
> > any means.
> >
> > Apache SOAP has never had any optimization effort put into
> > it. I am currently making some mods that'll improve things
> > a bit (someone found an obvious bottleneck with the way
> > SOAPMappingRegistry's were being created) but other than
> > that we never put any effort into making it fast. Axis is
> > charged with that mission .. Apache SOAP is a robust
> > implementation (given the pretty large number of downloads
> > its pretty impressive to not have any major bugs so far)
> > and not one that concentrated on performance.
> >
> > Sanjiva.
> >
> >
Re: web services performance
Posted by Sanjiva Weerawarana <sa...@watson.ibm.com>.
I certainly expect Axis to outperform to Apache SOAP. However,
comparing Axis and Apache SOAP right now is not useful (IMO)
until Axis supports SOAP Attachments too.
I believe GLUE does support SOAP Attachments.
Sanjiva.
----- Original Message -----
From: "Glen Daniels" <gd...@macromedia.com>
To: <so...@xml.apache.org>
Sent: Thursday, June 21, 2001 1:37 AM
Subject: Re: web services performance
> Just as an FYI, Axis on two of my machines benchmarked at about 4X and
about
> 8X the performance of ApacheSOAP 2.2 a few weeks ago (the 8X was certainly
> due to the fact that something in Axis (likely the parser) was able to
> utilize the dual processors on my work box to great effect). Still don't
> get no 500 msgs/second, though.... :) That said, Axis as it stands hasn't
> been seriously combed for optimizations either.
>
> --Glen
>
> ----- Original Message -----
> From: "Sanjiva Weerawarana" <sa...@watson.ibm.com>
> To: <so...@xml.apache.org>
> Sent: Thursday, June 21, 2001 12:01 AM
> Subject: Re: web services performance
>
>
> > > oh, so 5x wouldn't be good enough, huh? ;-)
> > >
> > > - graham
> >
> > :-) It certainly would be .. however, you'd have to put the
> > fastest possible JAXP parser with Apache SOAP and then try
> > it .. Xerces is solid and all, but ain't the fastest by
> > any means.
> >
> > Apache SOAP has never had any optimization effort put into
> > it. I am currently making some mods that'll improve things
> > a bit (someone found an obvious bottleneck with the way
> > SOAPMappingRegistry's were being created) but other than
> > that we never put any effort into making it fast. Axis is
> > charged with that mission .. Apache SOAP is a robust
> > implementation (given the pretty large number of downloads
> > its pretty impressive to not have any major bugs so far)
> > and not one that concentrated on performance.
> >
> > Sanjiva.
> >
> >
Re: web services performance
Posted by Glen Daniels <gd...@macromedia.com>.
Just as an FYI, Axis on two of my machines benchmarked at about 4X and about
8X the performance of ApacheSOAP 2.2 a few weeks ago (the 8X was certainly
due to the fact that something in Axis (likely the parser) was able to
utilize the dual processors on my work box to great effect). Still don't
get no 500 msgs/second, though.... :) That said, Axis as it stands hasn't
been seriously combed for optimizations either.
--Glen
----- Original Message -----
From: "Sanjiva Weerawarana" <sa...@watson.ibm.com>
To: <so...@xml.apache.org>
Sent: Thursday, June 21, 2001 12:01 AM
Subject: Re: web services performance
> > oh, so 5x wouldn't be good enough, huh? ;-)
> >
> > - graham
>
> :-) It certainly would be .. however, you'd have to put the
> fastest possible JAXP parser with Apache SOAP and then try
> it .. Xerces is solid and all, but ain't the fastest by
> any means.
>
> Apache SOAP has never had any optimization effort put into
> it. I am currently making some mods that'll improve things
> a bit (someone found an obvious bottleneck with the way
> SOAPMappingRegistry's were being created) but other than
> that we never put any effort into making it fast. Axis is
> charged with that mission .. Apache SOAP is a robust
> implementation (given the pretty large number of downloads
> its pretty impressive to not have any major bugs so far)
> and not one that concentrated on performance.
>
> Sanjiva.
>
>
Re: web services performance
Posted by Glen Daniels <gd...@macromedia.com>.
Just as an FYI, Axis on two of my machines benchmarked at about 4X and about
8X the performance of ApacheSOAP 2.2 a few weeks ago (the 8X was certainly
due to the fact that something in Axis (likely the parser) was able to
utilize the dual processors on my work box to great effect). Still don't
get no 500 msgs/second, though.... :) That said, Axis as it stands hasn't
been seriously combed for optimizations either.
--Glen
----- Original Message -----
From: "Sanjiva Weerawarana" <sa...@watson.ibm.com>
To: <so...@xml.apache.org>
Sent: Thursday, June 21, 2001 12:01 AM
Subject: Re: web services performance
> > oh, so 5x wouldn't be good enough, huh? ;-)
> >
> > - graham
>
> :-) It certainly would be .. however, you'd have to put the
> fastest possible JAXP parser with Apache SOAP and then try
> it .. Xerces is solid and all, but ain't the fastest by
> any means.
>
> Apache SOAP has never had any optimization effort put into
> it. I am currently making some mods that'll improve things
> a bit (someone found an obvious bottleneck with the way
> SOAPMappingRegistry's were being created) but other than
> that we never put any effort into making it fast. Axis is
> charged with that mission .. Apache SOAP is a robust
> implementation (given the pretty large number of downloads
> its pretty impressive to not have any major bugs so far)
> and not one that concentrated on performance.
>
> Sanjiva.
>
>
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
hi sanjiva,
any idea when Axis will be available? i'm writing a book
on web services for prentice hall and already have chapters
on J2EE/SOAP and .NET/SOAP. it would be great to include
a chapter on AXIS/SOAP!
cheers,
graham
-----Original Message-----
From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
Sent: Wednesday, June 20, 2001 11:02 PM
To: soap-user@xml.apache.org
Subject: Re: web services performance
> oh, so 5x wouldn't be good enough, huh? ;-)
>
> - graham
:-) It certainly would be .. however, you'd have to put the
fastest possible JAXP parser with Apache SOAP and then try
it .. Xerces is solid and all, but ain't the fastest by
any means.
Apache SOAP has never had any optimization effort put into
it. I am currently making some mods that'll improve things
a bit (someone found an obvious bottleneck with the way
SOAPMappingRegistry's were being created) but other than
that we never put any effort into making it fast. Axis is
charged with that mission .. Apache SOAP is a robust
implementation (given the pretty large number of downloads
its pretty impressive to not have any major bugs so far)
and not one that concentrated on performance.
Sanjiva.
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
hi sanjiva,
any idea when Axis will be available? i'm writing a book
on web services for prentice hall and already have chapters
on J2EE/SOAP and .NET/SOAP. it would be great to include
a chapter on AXIS/SOAP!
cheers,
graham
-----Original Message-----
From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
Sent: Wednesday, June 20, 2001 11:02 PM
To: soap-user@xml.apache.org
Subject: Re: web services performance
> oh, so 5x wouldn't be good enough, huh? ;-)
>
> - graham
:-) It certainly would be .. however, you'd have to put the
fastest possible JAXP parser with Apache SOAP and then try
it .. Xerces is solid and all, but ain't the fastest by
any means.
Apache SOAP has never had any optimization effort put into
it. I am currently making some mods that'll improve things
a bit (someone found an obvious bottleneck with the way
SOAPMappingRegistry's were being created) but other than
that we never put any effort into making it fast. Axis is
charged with that mission .. Apache SOAP is a robust
implementation (given the pretty large number of downloads
its pretty impressive to not have any major bugs so far)
and not one that concentrated on performance.
Sanjiva.
Re: web services performance
Posted by Sanjiva Weerawarana <sa...@watson.ibm.com>.
> oh, so 5x wouldn't be good enough, huh? ;-)
>
> - graham
:-) It certainly would be .. however, you'd have to put the
fastest possible JAXP parser with Apache SOAP and then try
it .. Xerces is solid and all, but ain't the fastest by
any means.
Apache SOAP has never had any optimization effort put into
it. I am currently making some mods that'll improve things
a bit (someone found an obvious bottleneck with the way
SOAPMappingRegistry's were being created) but other than
that we never put any effort into making it fast. Axis is
charged with that mission .. Apache SOAP is a robust
implementation (given the pretty large number of downloads
its pretty impressive to not have any major bugs so far)
and not one that concentrated on performance.
Sanjiva.
Re: web services performance
Posted by Sanjiva Weerawarana <sa...@watson.ibm.com>.
> oh, so 5x wouldn't be good enough, huh? ;-)
>
> - graham
:-) It certainly would be .. however, you'd have to put the
fastest possible JAXP parser with Apache SOAP and then try
it .. Xerces is solid and all, but ain't the fastest by
any means.
Apache SOAP has never had any optimization effort put into
it. I am currently making some mods that'll improve things
a bit (someone found an obvious bottleneck with the way
SOAPMappingRegistry's were being created) but other than
that we never put any effort into making it fast. Axis is
charged with that mission .. Apache SOAP is a robust
implementation (given the pretty large number of downloads
its pretty impressive to not have any major bugs so far)
and not one that concentrated on performance.
Sanjiva.
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
oh, so 5x wouldn't be good enough, huh? ;-)
- graham
-----Original Message-----
From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
Sent: Wednesday, June 20, 2001 10:23 PM
To: soap-user@xml.apache.org
Subject: Re: web services performance
On a similar note, it would be interesting to hold say the calling
end fixed (pick one- apache soap, glue, .net, whatever) and compare
the perf of glue vs. apache soap on the serving end. If that shows
glue an order of magnitude (10x) faster than apache soap then hats
off to you Graham!
Sanjiva.
----- Original Message -----
From: "Glen Daniels" <gd...@macromedia.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 11:07 PM
Subject: Re: web services performance
> Hi Graham!
>
> I'm curious - does your 500messages/second figure rely on "99.44%" SOAP,
or
> do you use special GLUE->GLUE patterns to speed things up? I.e. should
you
> expect that same rate (or similar) from GLUE->.NET, say?
>
> --Glen
>
> ----- Original Message -----
> From: "graham glass" <gr...@mindspring.com>
> To: <so...@xml.apache.org>
> Sent: Wednesday, June 20, 2001 10:29 PM
> Subject: RE: web services performance
>
>
> > hi sachin,
> >
> > please note that some web services platforms are
> > faster than others. just because RMI is 6x faster
> > than one particular implementation does not imply
> > that RMI is 6x faster than "web services".
> >
> > our own platform, "GLUE", is optimized for web
> > services and gets about 800 round trip messages
> > per second between two VMs running on the same
> > machine, and around 500 messages/second when
> > they're on different machines on a low speed LAN.
> >
> > cheers,
> > graham
> >
> > http://www.themindelectric.com
> >
> > -----Original Message-----
> > From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> > Sent: Wednesday, June 20, 2001 9:03 PM
> > To: soap user list
> > Subject: web services performance
> >
> >
> > Hi,
> >
> > I did a simple test to see how much performance
> > penalty one will have to pay for Openness of SOAP.
> > RMI was 6 times faster than web services. I used a
> > very simple function call to test this. (I am not
> > claiming that I am right, but this is what I got!!!)
> >
> > I am sure other architects are also facing the
> > challenge of deciding what level of services one needs
> > to expose as web services. If we decide to expose
> > fundamental services then performace of the
> > applications will be very slow.
> >
> > I am very interested in knowing what you think about
> > the future of web services performance? Any benchmark
> > data available to compare web services & other
> > protocals like CORBA or RMI or servlets.
> >
> > Any comments will be highly appreciated.
> >
> > S
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Get personalized email addresses from Yahoo! Mail - only $35
> > a year! http://personal.mail.yahoo.com/
> >
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
oh, so 5x wouldn't be good enough, huh? ;-)
- graham
-----Original Message-----
From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
Sent: Wednesday, June 20, 2001 10:23 PM
To: soap-user@xml.apache.org
Subject: Re: web services performance
On a similar note, it would be interesting to hold say the calling
end fixed (pick one- apache soap, glue, .net, whatever) and compare
the perf of glue vs. apache soap on the serving end. If that shows
glue an order of magnitude (10x) faster than apache soap then hats
off to you Graham!
Sanjiva.
----- Original Message -----
From: "Glen Daniels" <gd...@macromedia.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 11:07 PM
Subject: Re: web services performance
> Hi Graham!
>
> I'm curious - does your 500messages/second figure rely on "99.44%" SOAP,
or
> do you use special GLUE->GLUE patterns to speed things up? I.e. should
you
> expect that same rate (or similar) from GLUE->.NET, say?
>
> --Glen
>
> ----- Original Message -----
> From: "graham glass" <gr...@mindspring.com>
> To: <so...@xml.apache.org>
> Sent: Wednesday, June 20, 2001 10:29 PM
> Subject: RE: web services performance
>
>
> > hi sachin,
> >
> > please note that some web services platforms are
> > faster than others. just because RMI is 6x faster
> > than one particular implementation does not imply
> > that RMI is 6x faster than "web services".
> >
> > our own platform, "GLUE", is optimized for web
> > services and gets about 800 round trip messages
> > per second between two VMs running on the same
> > machine, and around 500 messages/second when
> > they're on different machines on a low speed LAN.
> >
> > cheers,
> > graham
> >
> > http://www.themindelectric.com
> >
> > -----Original Message-----
> > From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> > Sent: Wednesday, June 20, 2001 9:03 PM
> > To: soap user list
> > Subject: web services performance
> >
> >
> > Hi,
> >
> > I did a simple test to see how much performance
> > penalty one will have to pay for Openness of SOAP.
> > RMI was 6 times faster than web services. I used a
> > very simple function call to test this. (I am not
> > claiming that I am right, but this is what I got!!!)
> >
> > I am sure other architects are also facing the
> > challenge of deciding what level of services one needs
> > to expose as web services. If we decide to expose
> > fundamental services then performace of the
> > applications will be very slow.
> >
> > I am very interested in knowing what you think about
> > the future of web services performance? Any benchmark
> > data available to compare web services & other
> > protocals like CORBA or RMI or servlets.
> >
> > Any comments will be highly appreciated.
> >
> > S
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Get personalized email addresses from Yahoo! Mail - only $35
> > a year! http://personal.mail.yahoo.com/
> >
Re: web services performance
Posted by Sanjiva Weerawarana <sa...@watson.ibm.com>.
On a similar note, it would be interesting to hold say the calling
end fixed (pick one- apache soap, glue, .net, whatever) and compare
the perf of glue vs. apache soap on the serving end. If that shows
glue an order of magnitude (10x) faster than apache soap then hats
off to you Graham!
Sanjiva.
----- Original Message -----
From: "Glen Daniels" <gd...@macromedia.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 11:07 PM
Subject: Re: web services performance
> Hi Graham!
>
> I'm curious - does your 500messages/second figure rely on "99.44%" SOAP,
or
> do you use special GLUE->GLUE patterns to speed things up? I.e. should
you
> expect that same rate (or similar) from GLUE->.NET, say?
>
> --Glen
>
> ----- Original Message -----
> From: "graham glass" <gr...@mindspring.com>
> To: <so...@xml.apache.org>
> Sent: Wednesday, June 20, 2001 10:29 PM
> Subject: RE: web services performance
>
>
> > hi sachin,
> >
> > please note that some web services platforms are
> > faster than others. just because RMI is 6x faster
> > than one particular implementation does not imply
> > that RMI is 6x faster than "web services".
> >
> > our own platform, "GLUE", is optimized for web
> > services and gets about 800 round trip messages
> > per second between two VMs running on the same
> > machine, and around 500 messages/second when
> > they're on different machines on a low speed LAN.
> >
> > cheers,
> > graham
> >
> > http://www.themindelectric.com
> >
> > -----Original Message-----
> > From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> > Sent: Wednesday, June 20, 2001 9:03 PM
> > To: soap user list
> > Subject: web services performance
> >
> >
> > Hi,
> >
> > I did a simple test to see how much performance
> > penalty one will have to pay for Openness of SOAP.
> > RMI was 6 times faster than web services. I used a
> > very simple function call to test this. (I am not
> > claiming that I am right, but this is what I got!!!)
> >
> > I am sure other architects are also facing the
> > challenge of deciding what level of services one needs
> > to expose as web services. If we decide to expose
> > fundamental services then performace of the
> > applications will be very slow.
> >
> > I am very interested in knowing what you think about
> > the future of web services performance? Any benchmark
> > data available to compare web services & other
> > protocals like CORBA or RMI or servlets.
> >
> > Any comments will be highly appreciated.
> >
> > S
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Get personalized email addresses from Yahoo! Mail - only $35
> > a year! http://personal.mail.yahoo.com/
> >
Re: web services performance
Posted by Sanjiva Weerawarana <sa...@watson.ibm.com>.
On a similar note, it would be interesting to hold say the calling
end fixed (pick one- apache soap, glue, .net, whatever) and compare
the perf of glue vs. apache soap on the serving end. If that shows
glue an order of magnitude (10x) faster than apache soap then hats
off to you Graham!
Sanjiva.
----- Original Message -----
From: "Glen Daniels" <gd...@macromedia.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 11:07 PM
Subject: Re: web services performance
> Hi Graham!
>
> I'm curious - does your 500messages/second figure rely on "99.44%" SOAP,
or
> do you use special GLUE->GLUE patterns to speed things up? I.e. should
you
> expect that same rate (or similar) from GLUE->.NET, say?
>
> --Glen
>
> ----- Original Message -----
> From: "graham glass" <gr...@mindspring.com>
> To: <so...@xml.apache.org>
> Sent: Wednesday, June 20, 2001 10:29 PM
> Subject: RE: web services performance
>
>
> > hi sachin,
> >
> > please note that some web services platforms are
> > faster than others. just because RMI is 6x faster
> > than one particular implementation does not imply
> > that RMI is 6x faster than "web services".
> >
> > our own platform, "GLUE", is optimized for web
> > services and gets about 800 round trip messages
> > per second between two VMs running on the same
> > machine, and around 500 messages/second when
> > they're on different machines on a low speed LAN.
> >
> > cheers,
> > graham
> >
> > http://www.themindelectric.com
> >
> > -----Original Message-----
> > From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> > Sent: Wednesday, June 20, 2001 9:03 PM
> > To: soap user list
> > Subject: web services performance
> >
> >
> > Hi,
> >
> > I did a simple test to see how much performance
> > penalty one will have to pay for Openness of SOAP.
> > RMI was 6 times faster than web services. I used a
> > very simple function call to test this. (I am not
> > claiming that I am right, but this is what I got!!!)
> >
> > I am sure other architects are also facing the
> > challenge of deciding what level of services one needs
> > to expose as web services. If we decide to expose
> > fundamental services then performace of the
> > applications will be very slow.
> >
> > I am very interested in knowing what you think about
> > the future of web services performance? Any benchmark
> > data available to compare web services & other
> > protocals like CORBA or RMI or servlets.
> >
> > Any comments will be highly appreciated.
> >
> > S
> >
> > __________________________________________________
> > Do You Yahoo!?
> > Get personalized email addresses from Yahoo! Mail - only $35
> > a year! http://personal.mail.yahoo.com/
> >
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
hi glen,
only some of it is due to GLUE<->GLUE optimizations.
from what i've been told, GLUE<-><other> is pretty fast too.
i'll leave it to others to create the performance benchmarks! ;-)
cheers,
graham
p.s. GLUE 3.1, due out soon, is about 10% faster than 3.0
-----Original Message-----
From: Glen Daniels [mailto:gdaniels@macromedia.com]
Sent: Wednesday, June 20, 2001 10:07 PM
To: soap-user@xml.apache.org
Subject: Re: web services performance
Hi Graham!
I'm curious - does your 500messages/second figure rely on "99.44%" SOAP, or
do you use special GLUE->GLUE patterns to speed things up? I.e. should you
expect that same rate (or similar) from GLUE->.NET, say?
--Glen
----- Original Message -----
From: "graham glass" <gr...@mindspring.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 10:29 PM
Subject: RE: web services performance
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail - only $35
> a year! http://personal.mail.yahoo.com/
>
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
hi glen,
only some of it is due to GLUE<->GLUE optimizations.
from what i've been told, GLUE<-><other> is pretty fast too.
i'll leave it to others to create the performance benchmarks! ;-)
cheers,
graham
p.s. GLUE 3.1, due out soon, is about 10% faster than 3.0
-----Original Message-----
From: Glen Daniels [mailto:gdaniels@macromedia.com]
Sent: Wednesday, June 20, 2001 10:07 PM
To: soap-user@xml.apache.org
Subject: Re: web services performance
Hi Graham!
I'm curious - does your 500messages/second figure rely on "99.44%" SOAP, or
do you use special GLUE->GLUE patterns to speed things up? I.e. should you
expect that same rate (or similar) from GLUE->.NET, say?
--Glen
----- Original Message -----
From: "graham glass" <gr...@mindspring.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 10:29 PM
Subject: RE: web services performance
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail - only $35
> a year! http://personal.mail.yahoo.com/
>
Re: web services performance
Posted by Glen Daniels <gd...@macromedia.com>.
Hi Graham!
I'm curious - does your 500messages/second figure rely on "99.44%" SOAP, or
do you use special GLUE->GLUE patterns to speed things up? I.e. should you
expect that same rate (or similar) from GLUE->.NET, say?
--Glen
----- Original Message -----
From: "graham glass" <gr...@mindspring.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 10:29 PM
Subject: RE: web services performance
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail - only $35
> a year! http://personal.mail.yahoo.com/
>
Re: web services performance
Posted by Glen Daniels <gd...@macromedia.com>.
Hi Graham!
I'm curious - does your 500messages/second figure rely on "99.44%" SOAP, or
do you use special GLUE->GLUE patterns to speed things up? I.e. should you
expect that same rate (or similar) from GLUE->.NET, say?
--Glen
----- Original Message -----
From: "graham glass" <gr...@mindspring.com>
To: <so...@xml.apache.org>
Sent: Wednesday, June 20, 2001 10:29 PM
Subject: RE: web services performance
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail - only $35
> a year! http://personal.mail.yahoo.com/
>
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
yes, but no-one believes you when you spout your own
benchmarks, so i encourage you to perform your own ;-)
-----Original Message-----
From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
Sent: Wednesday, June 20, 2001 9:31 PM
To: soap-user@xml.apache.org
Subject: RE: web services performance
graham
This is really a good new. Have you compared GLUE with
RMI or other implementations?
--- graham glass <gr...@mindspring.com> wrote:
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari
> [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one
> needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any
> benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail -
> only $35
> a year! http://personal.mail.yahoo.com/
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
yes, but no-one believes you when you spout your own
benchmarks, so i encourage you to perform your own ;-)
-----Original Message-----
From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
Sent: Wednesday, June 20, 2001 9:31 PM
To: soap-user@xml.apache.org
Subject: RE: web services performance
graham
This is really a good new. Have you compared GLUE with
RMI or other implementations?
--- graham glass <gr...@mindspring.com> wrote:
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari
> [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one
> needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any
> benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail -
> only $35
> a year! http://personal.mail.yahoo.com/
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
RE: web services performance
Posted by sachin chaudhari <ch...@yahoo.com>.
graham
This is really a good new. Have you compared GLUE with
RMI or other implementations?
--- graham glass <gr...@mindspring.com> wrote:
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari
> [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one
> needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any
> benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail -
> only $35
> a year! http://personal.mail.yahoo.com/
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
RE: web services performance
Posted by sachin chaudhari <ch...@yahoo.com>.
graham
This is really a good new. Have you compared GLUE with
RMI or other implementations?
--- graham glass <gr...@mindspring.com> wrote:
> hi sachin,
>
> please note that some web services platforms are
> faster than others. just because RMI is 6x faster
> than one particular implementation does not imply
> that RMI is 6x faster than "web services".
>
> our own platform, "GLUE", is optimized for web
> services and gets about 800 round trip messages
> per second between two VMs running on the same
> machine, and around 500 messages/second when
> they're on different machines on a low speed LAN.
>
> cheers,
> graham
>
> http://www.themindelectric.com
>
> -----Original Message-----
> From: sachin chaudhari
> [mailto:chaudhsb2001@yahoo.com]
> Sent: Wednesday, June 20, 2001 9:03 PM
> To: soap user list
> Subject: web services performance
>
>
> Hi,
>
> I did a simple test to see how much performance
> penalty one will have to pay for Openness of SOAP.
> RMI was 6 times faster than web services. I used a
> very simple function call to test this. (I am not
> claiming that I am right, but this is what I got!!!)
>
> I am sure other architects are also facing the
> challenge of deciding what level of services one
> needs
> to expose as web services. If we decide to expose
> fundamental services then performace of the
> applications will be very slow.
>
> I am very interested in knowing what you think about
> the future of web services performance? Any
> benchmark
> data available to compare web services & other
> protocals like CORBA or RMI or servlets.
>
> Any comments will be highly appreciated.
>
> S
>
> __________________________________________________
> Do You Yahoo!?
> Get personalized email addresses from Yahoo! Mail -
> only $35
> a year! http://personal.mail.yahoo.com/
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
hi sachin,
please note that some web services platforms are
faster than others. just because RMI is 6x faster
than one particular implementation does not imply
that RMI is 6x faster than "web services".
our own platform, "GLUE", is optimized for web
services and gets about 800 round trip messages
per second between two VMs running on the same
machine, and around 500 messages/second when
they're on different machines on a low speed LAN.
cheers,
graham
http://www.themindelectric.com
-----Original Message-----
From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
Sent: Wednesday, June 20, 2001 9:03 PM
To: soap user list
Subject: web services performance
Hi,
I did a simple test to see how much performance
penalty one will have to pay for Openness of SOAP.
RMI was 6 times faster than web services. I used a
very simple function call to test this. (I am not
claiming that I am right, but this is what I got!!!)
I am sure other architects are also facing the
challenge of deciding what level of services one needs
to expose as web services. If we decide to expose
fundamental services then performace of the
applications will be very slow.
I am very interested in knowing what you think about
the future of web services performance? Any benchmark
data available to compare web services & other
protocals like CORBA or RMI or servlets.
Any comments will be highly appreciated.
S
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/
RE: web services performance
Posted by graham glass <gr...@mindspring.com>.
hi sachin,
please note that some web services platforms are
faster than others. just because RMI is 6x faster
than one particular implementation does not imply
that RMI is 6x faster than "web services".
our own platform, "GLUE", is optimized for web
services and gets about 800 round trip messages
per second between two VMs running on the same
machine, and around 500 messages/second when
they're on different machines on a low speed LAN.
cheers,
graham
http://www.themindelectric.com
-----Original Message-----
From: sachin chaudhari [mailto:chaudhsb2001@yahoo.com]
Sent: Wednesday, June 20, 2001 9:03 PM
To: soap user list
Subject: web services performance
Hi,
I did a simple test to see how much performance
penalty one will have to pay for Openness of SOAP.
RMI was 6 times faster than web services. I used a
very simple function call to test this. (I am not
claiming that I am right, but this is what I got!!!)
I am sure other architects are also facing the
challenge of deciding what level of services one needs
to expose as web services. If we decide to expose
fundamental services then performace of the
applications will be very slow.
I am very interested in knowing what you think about
the future of web services performance? Any benchmark
data available to compare web services & other
protocals like CORBA or RMI or servlets.
Any comments will be highly appreciated.
S
__________________________________________________
Do You Yahoo!?
Get personalized email addresses from Yahoo! Mail - only $35
a year! http://personal.mail.yahoo.com/