jmeter-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Shmuel Krakower <shmul...@gmail.com>
Subject Re: Implementation ideas for core changes to ease implementation of Websocket, HTTP2, SPDY....
Date Tue, 16 Dec 2014 08:18:35 GMT
Higher level concerns / thoughts.

First:
Those implementations make the browser support "incoming" requests thus
effectively making the browser to be able to serve/act/respond, as a
"server" for a single (or very few) open connections.

I wonder if trying to support those protocols will not put JMeter into a
situation which it actually acts as a "server" (listening to thousands of
open connections).
If you think of the situation today, it may happen that you will have tens
or hundreds of parallel open connections in any given time, but you, as the
creator of the load test, have control of that as you define the scenario.

With those implementations, JMeter (and the testser) is no longer in
control of how many open connections there are at any given time, as well
as not in control of the incoming data throughput which JMeter need to deal
with.
I wonder if that will not bring JMeter to be the bottleneck of the test in
more cases of as today? Mostly because its role is now getting more complex?

Second thought:
What should those new samplers (or single sampler) should measure? With
normal sync requests the response time is the measurement, but in async
request, with potentially multiple / unlimited responses, what should be
measured there?


Shmuel Krakower.
www.Beatsoo.org - re-use your jmeter scripts for application performance
monitoring from worldwide locations for free.

On Tue, Dec 16, 2014 at 3:58 AM, sebb <sebbaz@gmail.com> wrote:
>
> On 15 December 2014 at 19:40, Philippe Mouawad
> <philippe.mouawad@gmail.com> wrote:
> > Hello,
> > By this thread I would like to start a discussion on core changes we
> should
> > implement in JMeter to easily implement new protocols like Websocket,
> HTTP
> > 2, SPDY.
> > For recap of some of those protocols:
> > -
> >
> http://stackoverflow.com/questions/11077857/what-are-long-polling-websockets-server-sent-events-sse-and-comet
> >
> > Some of the main problems I see:
> >
> >    - It seems we need to use Async IO (
> >    https://hc.apache.org/httpcomponents-asyncclient-dev/quickstart.html)
> as
> >    to scale we cannot afford holding thread until we get a response
>
> The advice generally given on the HC mailing lists is that Asynch IO
> only starts to show benefits when handling 1000s of threads.
>
> >    - In this case assertions are to be made on result that we cannot get
> >    immediately. So Assertions would become FutureAssertion , I rather
> see how
> >    to do ths
> >    - But what about extractors ? How would they be used ? Experience in
> >    those use cases are welcome
> >
> >
> > Regards
> > @philmdot
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message