jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mike Stover" <mstov...@apache.org>
Subject Re: How accurate is JMeter?
Date Thu, 20 Jun 2002 02:32:20 GMT
If you're getting instanteous response times regardless of the number of users, you need to

verify that the server is returning valid results.  Server error, or errors in your test script
can 
mean response times are very quick - simply because nothing is really happening.

Run your test with just one user, use the view results tree listener and examine the response
to 
each request - make sure it's right. 

Another thing to do is, while one machine is running a big test with many users, check the

validity of the responses on another machine, using JMeter running just one user - again,

watch the response carefully and make sure everything is working right.  

You can also use assertions to verify that the html returned by the server is correct.  Some
of 
the visualizers will give an indication of whether the request resulted in an error or not
(and will 
use the assertions you specify to determine failure or success).  The most recent milestone

release (1.7.1) provides more functionality in this area.

And lastly, you can manually go through a simple test plan in a browser while JMeter bangs

the server, and see for yourself, subjectively, how fast the response times are.  Thus, you
can 
see for yourself how the server is responding to you, while also responding to 200 other 
users.

If you server is returning correct results, the times JMeter reports are only going to error
by 
being longer than the real times.  JMeter will never report a response time that is smaller
than 
the actual time was.  But you do have to watch out for errors, because they often return very

quickly (ie 404 errors are usually just a static page, so it's very fast).

-Mike

On 20 Jun 2002 at 11:14, De Gene Ng wrote:

> To Whom It May Concern,
>  
>  I have been using the Jakarta JMeter application to load test a website for
>  a client and I have managed to obtain response times for up to 200 users.
> I
>  have plotted these results on a graph and the graph is increasing linearly
>  as expected.  However, how can I verify that those results/response times
>  determined by JMeter are accurate?  Even with 200 simultaneous users
>  (ramp-up period of 0 secs), why do I still obtain an almost-instantaneous
>  minimum response time (20 milliseconds)?
>  
>  Thank you for your time,
>  
>  Regards,
>  
>  De Gene Ng
>  Aspect Computing P/L
>  551 Glenferrie Road
>  Hawthorn
>  Victoria 3122
>  AUSTRALIA
>  Tel: (+61 3) 9230 2222
>  Mob: 0403 432 169
>  
> 
> 
> Andy Redwood wrote:
>  
>  To Whom It May Concern,
>  
>  Just to extend the first query below a little bit ie. "how can I verify
> that
>  those results/response times determined by JMeter are accurate?":
>  
>  Jakarta JMeter simulates X users (in our case up to 220 users)
>  simultaneously hitting a Web server to get a page delivered.  By what
>  mechanism does JMeter simulate the "real world" .... ie. if JMeter comes up
>  with a 5 sec average response for 220 users, how do we have confidence that
>  this is going to be similar to the real-world response when 220 users hit
>  that Web server?
>  
>  Thanks and Regards,
>  
>  Andy
> 
> 
> 
> ************************************************************************
> MIMEsweeper has been used to check this email for security
> ************************************************************************
> 
> 
> --
> To unsubscribe, e-mail:   <mailto:jmeter-user-unsubscribe@jakarta.apache.org>
> For additional commands, e-mail: <mailto:jmeter-user-help@jakarta.apache.org>
> 



--
Michael Stover
mstover1@apache.org
Yahoo IM: mstover_ya
ICQ: 152975688

--
To unsubscribe, e-mail:   <mailto:jmeter-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:jmeter-user-help@jakarta.apache.org>


Mime
View raw message