You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by DELHOSTE Fabrice <Fa...@gemalto.com> on 2007/10/19 20:35:35 UTC
Asynchronous samplers
Hi,
I am currently trying to design a custom sampler upon an asynchronous
request-response API and I haven't found enough detailed discussion on
that topic.
I basically want to measure the time from sending the request to its
received response.
As far as I understand, there may be a need for even-driven sampler in
the future but what's the best way to do it today?
I have first tried to add subresults on the SampleResult of each request
when receiving responses (in another thread) without success.
Tree of results were not correctly updated and I haven't found the way
to publish them as new SampleResult instead of subresult.
So I'm thinking a bit more now :-) ... about separating completely the
request sampler from the response sampler.
I would typically have 2 thread groups, one for request, the other for
response.
The request sampler would send its request, not waiting for anything.
The response sampler would listen for a response until some is received
(using a shared lock with the listener initiated in request sampler),
then return it as SampleResult.
In theory, I would obtain on one side, all my request data, and on the
other side, my response data, that I would be able to correlate with my
request...and then proceed naturally with all JMeter powerful stuff on
both groups.
Before starting, I'd like to know if this seems possible to you.
Any advice, or another way to implement?
Thanks,
Fabrice
Re: Asynchronous samplers
Posted by sebb <se...@gmail.com>.
The number of client threads is surely irrelevant to the application under test?
What's important is the number of requests and the rate at which they arrive.
You just need to ensure that the application thinks that the requests
come from the approriate source.
This is straying away from what should be discussed on the JMeter user list.
JMeter is focussed on testing request-reply protocols.
On 22/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
>
> :-)
>
> Ok, shoot again: what if 20,000 requests?
> Moreover, having 200 threads would not simulate exactly my use case
> (just 3 threads in my own client application).
>
> I'm currently trying the thing with 2 separated samplers, I'll give you
> feedback.
>
>
> -----Original Message-----
> From: sebb [mailto:sebbaz@gmail.com]
> Sent: lundi 22 octobre 2007 11:41
> To: JMeter Users List
> Subject: Re: Asynchronous samplers
>
> On 22/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> > Thanks for your answer.
> >
> > > If there is always a response, why not send the request, and then
> > > wait
> > for the response?
> >
> > Unfortunately, that would limit my concurrent number of requests to
> > the number of sampling threads. And I want much more.
> > Ex: what if I want to simulate 200 concurrent requests made by 3
> > client threads?
>
> Why not use 200 threads?
>
> > That's why I'm thinking about to make my request sampler not waiting
> > for response, returning a SampleResult related to the sending of
> request.
> > At the same time, I can build another SampleResult related to the
> > response that I just start in this sampler.
> > Then, a new "response" sampler could just wait for responses and
> > ending the SampleResult initiated in the request sample, for instance,
>
> > by sharing a Map of SampleResult as a variable between them.
> >
> > I've actually seen this kind of discussion in the history of the
> > mailing-list but there has been no follow-up...
> >
> > Fabrice
> >
> >
> > -----Original Message-----
> > From: sebb [mailto:sebbaz@gmail.com]
> > Sent: samedi 20 octobre 2007 01:48
> > To: JMeter Users List
> > Subject: Re: Asynchronous samplers
> >
> > On 19/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> > > Hi,
> > >
> > > I am currently trying to design a custom sampler upon an
> > > asynchronous request-response API and I haven't found enough
> > > detailed discussion on
> >
> > > that topic.
> >
> > JMeter does not support asynchronous working per se.
> >
> > > I basically want to measure the time from sending the request to its
>
> > > received response.
> >
> > If there is always a response, why not send the request, and then wait
>
> > for the response?
> >
> > This can be done in a single sampler.
> >
> > > As far as I understand, there may be a need for even-driven sampler
> > > in
> >
> > > the future but what's the best way to do it today?
> >
> > We have no plans to implement event-driven samplers.
> >
> > > I have first tried to add subresults on the SampleResult of each
> > > request when receiving responses (in another thread) without
> success.
> > > Tree of results were not correctly updated and I haven't found the
> > > way
> >
> > > to publish them as new SampleResult instead of subresult.
> > >
> > > So I'm thinking a bit more now :-) ... about separating completely
> > > the
> >
> > > request sampler from the response sampler.
> > >
> > > I would typically have 2 thread groups, one for request, the other
> > > for
> >
> > > response.
> > > The request sampler would send its request, not waiting for
> anything.
> > > The response sampler would listen for a response until some is
> > > received (using a shared lock with the listener initiated in request
>
> > > sampler), then return it as SampleResult.
> >
> > I would just send the request and wait for the response.
> >
> > Everything else would be a lot more work.
> >
> > > In theory, I would obtain on one side, all my request data, and on
> > > the
> >
> > > other side, my response data, that I would be able to correlate with
>
> > > my request...and then proceed naturally with all JMeter powerful
> > > stuff
> >
> > > on both groups.
> > >
> > > Before starting, I'd like to know if this seems possible to you.
> > > Any advice, or another way to implement?
> > >
> > > Thanks,
> > > Fabrice
> > >
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> >
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
RE: Asynchronous samplers
Posted by DELHOSTE Fabrice <Fa...@gemalto.com>.
:-)
Ok, shoot again: what if 20,000 requests?
Moreover, having 200 threads would not simulate exactly my use case
(just 3 threads in my own client application).
I'm currently trying the thing with 2 separated samplers, I'll give you
feedback.
-----Original Message-----
From: sebb [mailto:sebbaz@gmail.com]
Sent: lundi 22 octobre 2007 11:41
To: JMeter Users List
Subject: Re: Asynchronous samplers
On 22/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> Thanks for your answer.
>
> > If there is always a response, why not send the request, and then
> > wait
> for the response?
>
> Unfortunately, that would limit my concurrent number of requests to
> the number of sampling threads. And I want much more.
> Ex: what if I want to simulate 200 concurrent requests made by 3
> client threads?
Why not use 200 threads?
> That's why I'm thinking about to make my request sampler not waiting
> for response, returning a SampleResult related to the sending of
request.
> At the same time, I can build another SampleResult related to the
> response that I just start in this sampler.
> Then, a new "response" sampler could just wait for responses and
> ending the SampleResult initiated in the request sample, for instance,
> by sharing a Map of SampleResult as a variable between them.
>
> I've actually seen this kind of discussion in the history of the
> mailing-list but there has been no follow-up...
>
> Fabrice
>
>
> -----Original Message-----
> From: sebb [mailto:sebbaz@gmail.com]
> Sent: samedi 20 octobre 2007 01:48
> To: JMeter Users List
> Subject: Re: Asynchronous samplers
>
> On 19/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> > Hi,
> >
> > I am currently trying to design a custom sampler upon an
> > asynchronous request-response API and I haven't found enough
> > detailed discussion on
>
> > that topic.
>
> JMeter does not support asynchronous working per se.
>
> > I basically want to measure the time from sending the request to its
> > received response.
>
> If there is always a response, why not send the request, and then wait
> for the response?
>
> This can be done in a single sampler.
>
> > As far as I understand, there may be a need for even-driven sampler
> > in
>
> > the future but what's the best way to do it today?
>
> We have no plans to implement event-driven samplers.
>
> > I have first tried to add subresults on the SampleResult of each
> > request when receiving responses (in another thread) without
success.
> > Tree of results were not correctly updated and I haven't found the
> > way
>
> > to publish them as new SampleResult instead of subresult.
> >
> > So I'm thinking a bit more now :-) ... about separating completely
> > the
>
> > request sampler from the response sampler.
> >
> > I would typically have 2 thread groups, one for request, the other
> > for
>
> > response.
> > The request sampler would send its request, not waiting for
anything.
> > The response sampler would listen for a response until some is
> > received (using a shared lock with the listener initiated in request
> > sampler), then return it as SampleResult.
>
> I would just send the request and wait for the response.
>
> Everything else would be a lot more work.
>
> > In theory, I would obtain on one side, all my request data, and on
> > the
>
> > other side, my response data, that I would be able to correlate with
> > my request...and then proceed naturally with all JMeter powerful
> > stuff
>
> > on both groups.
> >
> > Before starting, I'd like to know if this seems possible to you.
> > Any advice, or another way to implement?
> >
> > Thanks,
> > Fabrice
> >
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Asynchronous samplers
Posted by sebb <se...@gmail.com>.
On 22/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> Thanks for your answer.
>
> > If there is always a response, why not send the request, and then wait
> for the response?
>
> Unfortunately, that would limit my concurrent number of requests to the
> number of sampling threads. And I want much more.
> Ex: what if I want to simulate 200 concurrent requests made by 3 client
> threads?
Why not use 200 threads?
> That's why I'm thinking about to make my request sampler not waiting for
> response, returning a SampleResult related to the sending of request.
> At the same time, I can build another SampleResult related to the
> response that I just start in this sampler.
> Then, a new "response" sampler could just wait for responses and ending
> the SampleResult initiated in the request sample, for instance, by
> sharing a Map of SampleResult as a variable between them.
>
> I've actually seen this kind of discussion in the history of the
> mailing-list but there has been no follow-up...
>
> Fabrice
>
>
> -----Original Message-----
> From: sebb [mailto:sebbaz@gmail.com]
> Sent: samedi 20 octobre 2007 01:48
> To: JMeter Users List
> Subject: Re: Asynchronous samplers
>
> On 19/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> > Hi,
> >
> > I am currently trying to design a custom sampler upon an asynchronous
> > request-response API and I haven't found enough detailed discussion on
>
> > that topic.
>
> JMeter does not support asynchronous working per se.
>
> > I basically want to measure the time from sending the request to its
> > received response.
>
> If there is always a response, why not send the request, and then wait
> for the response?
>
> This can be done in a single sampler.
>
> > As far as I understand, there may be a need for even-driven sampler in
>
> > the future but what's the best way to do it today?
>
> We have no plans to implement event-driven samplers.
>
> > I have first tried to add subresults on the SampleResult of each
> > request when receiving responses (in another thread) without success.
> > Tree of results were not correctly updated and I haven't found the way
>
> > to publish them as new SampleResult instead of subresult.
> >
> > So I'm thinking a bit more now :-) ... about separating completely the
>
> > request sampler from the response sampler.
> >
> > I would typically have 2 thread groups, one for request, the other for
>
> > response.
> > The request sampler would send its request, not waiting for anything.
> > The response sampler would listen for a response until some is
> > received (using a shared lock with the listener initiated in request
> > sampler), then return it as SampleResult.
>
> I would just send the request and wait for the response.
>
> Everything else would be a lot more work.
>
> > In theory, I would obtain on one side, all my request data, and on the
>
> > other side, my response data, that I would be able to correlate with
> > my request...and then proceed naturally with all JMeter powerful stuff
>
> > on both groups.
> >
> > Before starting, I'd like to know if this seems possible to you.
> > Any advice, or another way to implement?
> >
> > Thanks,
> > Fabrice
> >
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
RE: Asynchronous samplers
Posted by DELHOSTE Fabrice <Fa...@gemalto.com>.
Thanks for your answer.
> If there is always a response, why not send the request, and then wait
for the response?
Unfortunately, that would limit my concurrent number of requests to the
number of sampling threads. And I want much more.
Ex: what if I want to simulate 200 concurrent requests made by 3 client
threads?
That's why I'm thinking about to make my request sampler not waiting for
response, returning a SampleResult related to the sending of request.
At the same time, I can build another SampleResult related to the
response that I just start in this sampler.
Then, a new "response" sampler could just wait for responses and ending
the SampleResult initiated in the request sample, for instance, by
sharing a Map of SampleResult as a variable between them.
I've actually seen this kind of discussion in the history of the
mailing-list but there has been no follow-up...
Fabrice
-----Original Message-----
From: sebb [mailto:sebbaz@gmail.com]
Sent: samedi 20 octobre 2007 01:48
To: JMeter Users List
Subject: Re: Asynchronous samplers
On 19/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> Hi,
>
> I am currently trying to design a custom sampler upon an asynchronous
> request-response API and I haven't found enough detailed discussion on
> that topic.
JMeter does not support asynchronous working per se.
> I basically want to measure the time from sending the request to its
> received response.
If there is always a response, why not send the request, and then wait
for the response?
This can be done in a single sampler.
> As far as I understand, there may be a need for even-driven sampler in
> the future but what's the best way to do it today?
We have no plans to implement event-driven samplers.
> I have first tried to add subresults on the SampleResult of each
> request when receiving responses (in another thread) without success.
> Tree of results were not correctly updated and I haven't found the way
> to publish them as new SampleResult instead of subresult.
>
> So I'm thinking a bit more now :-) ... about separating completely the
> request sampler from the response sampler.
>
> I would typically have 2 thread groups, one for request, the other for
> response.
> The request sampler would send its request, not waiting for anything.
> The response sampler would listen for a response until some is
> received (using a shared lock with the listener initiated in request
> sampler), then return it as SampleResult.
I would just send the request and wait for the response.
Everything else would be a lot more work.
> In theory, I would obtain on one side, all my request data, and on the
> other side, my response data, that I would be able to correlate with
> my request...and then proceed naturally with all JMeter powerful stuff
> on both groups.
>
> Before starting, I'd like to know if this seems possible to you.
> Any advice, or another way to implement?
>
> Thanks,
> Fabrice
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Asynchronous samplers
Posted by sebb <se...@gmail.com>.
On 19/10/2007, DELHOSTE Fabrice <Fa...@gemalto.com> wrote:
> Hi,
>
> I am currently trying to design a custom sampler upon an asynchronous
> request-response API and I haven't found enough detailed discussion on
> that topic.
JMeter does not support asynchronous working per se.
> I basically want to measure the time from sending the request to its
> received response.
If there is always a response, why not send the request, and then wait
for the response?
This can be done in a single sampler.
> As far as I understand, there may be a need for even-driven sampler in
> the future but what's the best way to do it today?
We have no plans to implement event-driven samplers.
> I have first tried to add subresults on the SampleResult of each request
> when receiving responses (in another thread) without success.
> Tree of results were not correctly updated and I haven't found the way
> to publish them as new SampleResult instead of subresult.
>
> So I'm thinking a bit more now :-) ... about separating completely the
> request sampler from the response sampler.
>
> I would typically have 2 thread groups, one for request, the other for
> response.
> The request sampler would send its request, not waiting for anything.
> The response sampler would listen for a response until some is received
> (using a shared lock with the listener initiated in request sampler),
> then return it as SampleResult.
I would just send the request and wait for the response.
Everything else would be a lot more work.
> In theory, I would obtain on one side, all my request data, and on the
> other side, my response data, that I would be able to correlate with my
> request...and then proceed naturally with all JMeter powerful stuff on
> both groups.
>
> Before starting, I'd like to know if this seems possible to you.
> Any advice, or another way to implement?
>
> Thanks,
> Fabrice
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org