jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brett Cave <brettc...@gmail.com>
Subject Re: Rampup giving strange results
Date Fri, 05 Mar 2010 14:06:28 GMT
On Fri, Mar 5, 2010 at 4:00 PM, Deepak Goel <deicool@gmail.com> wrote:

> Bret
>
> Why should the threads come to end end before the next loop enters? They
> should remain and keep on executing the next loops. Once the rampup period
> has reached, the threads should remain constant executing the next loops.
>

I thought this would be 1 of 2 methods for determining when to loop.

Method 1.
Wait for all threads to end and then start next loop

Method 2.
Wait until rampup period has expired and then start next loop.


or is another method used?


> Deepak
>
> On Fri, Mar 5, 2010 at 7:14 PM, Brett Cave <brettcave@gmail.com> wrote:
>
> > On Tue, Mar 2, 2010 at 3:36 PM, Deepak Goel <deicool@gmail.com> wrote:
> >
> > > Hey Bret
> > >
> > > You might like to give a very large number in the number of loops
> instead
> > > of
> > > 1. That will ensure that all the threads are running concurrently at
> the
> > > end
> > > of the test.
> > >
> >
> > For each loop, do all running threads come to an end before the next loop
> > enters, or does the next batch of threads get started up once the rampup
> > period has been reached?
> >
> >
> > > Regards
> > > Deepak
> > >
> > > On Tue, Mar 2, 2010 at 5:15 PM, Brett Cave <brettcave@gmail.com>
> wrote:
> > >
> > > > On 3/2/10, Deepak Goel <deicool@gmail.com> wrote:
> > > > > Hey Bret
> > > > >
> > > > > How many page request have you given as input?
> > > > >
> > > > > Can you also please post the results in terms of number of
> > request/sec
> > > > > (Throughput) instead of total number of samples per page (request
> per
> > > > item)?
> > > > > (Jmeter still might be increasing the number of threads at the rate
> > of
> > > > 1.6
> > > > > per minute)
> > > >
> > > > Hi Deepak,
> > > >
> > > > I have checked out the source, it seems that this is not how the
> > > > ramp-up feature works. It works as follows:
> > > >
> > > > startDelay = ramp-up time / number of threads.
> > > > each thread is then started with threadNumber x startDelay set as its
> > > > start delay.
> > > >
> > > > For example, if 100 threads are configured for 600 seconds, then
> start
> > > > delay is 6.
> > > > All 100 threads will be initialized up front. The first thread will
> > > > have a start delay of 0, the 2nd will start after 6 seconds, then 3rd
> > > > after 12s, 4th after 18s and so on.
> > > >
> > > > This means that ramp-up threads is not an indication of the final
> > > > concurrent threads that will be run during the test plan, but rather
> > > > the total number of threads to run during the test plan.
> > > >
> > > > Obviously, working the maths backwards doesnt work well, as to test
> > > > specific levels of concurrency you need to know how long each thread
> > > > would last. There is also no limit of how many concurrent threads are
> > > > running at any time.
> > > >
> > > > I am working on a patch that would allow a more logical approach to
> > > > testing, where a flag can be set to indicate that the ramp-up threads
> > > > is a "final concurrent # of threads" rather than "total # of threads"
> > > > metric.
> > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > >
> > > > > Regards
> > > > > Deepak
> > > > >
> > > > > On Tue, Mar 2, 2010 at 12:54 PM, Brett Cave <brettcave@gmail.com>
> > > wrote:
> > > > >
> > > > >> Hi all,
> > > > >>
> > > > >> I have a test plan that I use to increase the load on a target
> > system.
> > > > >> There
> > > > >> is 1 thread group, configured as follows:
> > > > >> Threads: 200
> > > > >> Ramp-up Period: 7200
> > > > >> Loop count: 1
> > > > >>
> > > > >> To my understanding, this should increase the number of threads
> over
> > 2
> > > > >> hours
> > > > >> from 1 to 200 (about 1.6 more threads per minute).
> > > > >>
> > > > >> The test runs non-interactively and logs to a file. If I check
the
> > > logs
> > > > >> after the test, I see a total # of samples per page (requests
per
> > > item)
> > > > of
> > > > >> 100. I think my understanding of the ramp-up mode is incorrect
-
> > > should
> > > > >> JMeter not start with 2 threads, then increase upwards until
200
> > > threads
> > > > >> are
> > > > >> running at the end of the period? (i.e. 100 threads after 1 hour).
> > > This
> > > > >> should also result in a lot more samples that what I am getting.
> > > > >>
> > > > >> Regards,
> > > > >> Brett
> > > > >>
> > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Guten Tag
> > > > >
> > > > >
> > > > >   --
> > > > > Keigu
> > > > >
> > > > > Deepak
> > > > > +91-9765089593
> > > > > deicool@gmail.com
> > > > >
> > > > > Skype: thumsupdeicool
> > > > > Google talk: deicool
> > > > > Blog: http://loveandfearless.wordpress.com
> > > > >
> > > > > Check out my Work at:
> > > > > LinkedIn: http://in.linkedin.com/in/thumsupdeicool
> > > > >
> > > > > "Contribute to the world, environment and more :
> > > > http://www.gridrepublic.org
> > > > > "
> > > > >
> > > >
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
> > > > For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
> > > >
> > > >
> > >
> > >
> > > --
> > > Guten Tag
> > >
> > >
> > >  --
> > > Keigu
> > >
> > > Deepak
> > > +91-9765089593
> > > deicool@gmail.com
> > >
> > > Skype: thumsupdeicool
> > > Google talk: deicool
> > > Blog: http://loveandfearless.wordpress.com
> > >
> > > Check out my Work at:
> > > LinkedIn: http://in.linkedin.com/in/thumsupdeicool
> > >
> > > "Contribute to the world, environment and more :
> > > http://www.gridrepublic.org
> > > "
> > >
> >
>
>
>
> --
> Guten Tag
>
>
>  --
> Keigu
>
> Deepak
> +91-9765089593
> deicool@gmail.com
>
> Skype: thumsupdeicool
> Google talk: deicool
> Blog: http://loveandfearless.wordpress.com
>
> Check out my Work at:
> LinkedIn: http://in.linkedin.com/in/thumsupdeicool
>
> "Contribute to the world, environment and more :
> http://www.gridrepublic.org
> "
>

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