You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@jmeter.apache.org by David Heath <da...@davidheath.org> on 2008/09/26 10:39:00 UTC
Recommended parameters for modelling user request behaviour
Hi,
Hello I'm new to JMeter and this mailing list so firstly, thanks for a
great tool! I'm a web software developer and I'd like to use JMeter to
load test a web application I'm developing.
I have set up a test which is intended to model a representative
sample of user behaviour. The test performs a user login and various
common actions. I want to use this to answer the question "what is the
mean response time at a given level of concurrent users?". To this end
I ran the test repeatedly with thread numbers from 2, 4, 8, 16, 32,
64, and then looked at the results.
The answer I got was that at loads of 16 or more threads I was getting
response times of 6 seconds or more. This is clearly not acceptable
for any application. So, before I start digging into my code to try to
find out what's going on, I wanted to sanity check that my model is
reasonable. One area that I'm not sure about is the introduction of a
random delay between user requests. I did this by adding a Guassian
Random Timer as a child of the root Thread Group. If understood
correctly, this random delay will get applied to all HTTP samplers in
the tree. I used a mean of 700ms and standard deviation of 400ms.
So, I have two questions:
1) have I understood correctly that the timer gets applied to the
whole tree .. I don't need to add one for each HTTP sampler?
2) do the figures of mean 700ms and stdev 400ms seem reasonable? I
wonder if it's a bit too short?
Thanks in advance for any advice you can give.
David Heath
http://davidheath.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Recommended parameters for modelling user request behaviour
Posted by David Heath <da...@davidheath.org>.
Hi Sebb,
Thanks for your reply.
>> 2) do the figures of mean 700ms and stdev 400ms seem reasonable? I
>> wonder if it's a bit too short?
>
> What are the users likely to do?
> That depends on how much info is in your pages.
Lets say it's something of a similar complexity to amazon, so a user
may be searching for something to act on, or they may be browsing
things to find something to act on. I don't really have hard data for
user behaviour because the application is still under development and
we have only had a small number of invited users to test it. So, I'm
trying to predict the future a bit here which is why I'm interested to
know how others have modeled this.
How should I go about choosing suitable values? Is the guassian random
timer the best thing to use? What parameters do you use?
Thanks
David
---------------------------------------------------------------------
To unsubscribe, e-mail: jmeter-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: jmeter-user-help@jakarta.apache.org
Re: Recommended parameters for modelling user request behaviour
Posted by sebb <se...@gmail.com>.
On 26/09/2008, David Heath <da...@davidheath.org> wrote:
> Hi,
>
> Hello I'm new to JMeter and this mailing list so firstly, thanks for a
> great tool! I'm a web software developer and I'd like to use JMeter to
> load test a web application I'm developing.
>
> I have set up a test which is intended to model a representative
> sample of user behaviour. The test performs a user login and various
> common actions. I want to use this to answer the question "what is the
> mean response time at a given level of concurrent users?". To this end
> I ran the test repeatedly with thread numbers from 2, 4, 8, 16, 32,
> 64, and then looked at the results.
>
> The answer I got was that at loads of 16 or more threads I was getting
> response times of 6 seconds or more. This is clearly not acceptable
> for any application. So, before I start digging into my code to try to
> find out what's going on, I wanted to sanity check that my model is
> reasonable. One area that I'm not sure about is the introduction of a
> random delay between user requests. I did this by adding a Guassian
> Random Timer as a child of the root Thread Group. If understood
> correctly, this random delay will get applied to all HTTP samplers in
> the tree. I used a mean of 700ms and standard deviation of 400ms.
>
> So, I have two questions:
> 1) have I understood correctly that the timer gets applied to the
> whole tree .. I don't need to add one for each HTTP sampler?
Yes. No.
> 2) do the figures of mean 700ms and stdev 400ms seem reasonable? I
> wonder if it's a bit too short?
What are the users likely to do?
That depends on how much info is in your pages.
> Thanks in advance for any advice you can give.
>
> David Heath
> http://davidheath.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