Yes. Thanks to Deepak for a very good post. Also, Jörg, in your case it's even more simple mathematics: Case 1: Each thread does a request each ~30.7secs Case 2: Each thread does a request each ~19.6secs the intervals being avg. Timer pause plus avg. response time.
So, even though your latency increased tenfold, the average interval is still shorter in comparison. Cheers, Felix On 07/27/2010 10:42 PM, Deepak Goel wrote: > Hey > > Namaskara~Nalama~Guten Tag > > As you increase the load, both the throughput and the response time > increase. There is a knee beyond which throughput would remain constant and > the response time would then increase exponentially. > > Looks like you haven't reached that knee yet at 3.1 req/sec throughput, so > the throughput will increase to 5 req/sec (knee throughput). > > There are lot of other operations which this test is doing like network > time, client rendering time, server socket connections that might also add > to the increasing response time in addition to the time taken by your > application. > > Also as the response time has increased by 10 times, the throughput has only > increased less than 2 times (not much as compared to response time). > > You can draw the knee curve for both throughput against load, response time > over load if you can measure more test. > > Something like: > > Request/sec Response Time Throughput > > 700 3000 ms 3/sec > 5000 40000ms 4/sec > 9600 114300ms 5/sec (Almost the knee i think) > 15000 3337777ms 5.2/sec > 30000 7778888ms 5.3/sec > > If you put your results (the above are only a sample) in an excel sheet, and > get the curve chart, you would be able to see the response time curve > (exponential) and throughput curve(plateau over a hill) > > Deepak > -- > Keigu > > Deepak > +91-9765089593 > deic...@gmail.com > > Skype: thumsupdeicool > Google talk: deicool > Blog: http://loveandfearless.wordpress.com > Facebook: http://www.facebook.com/deicool > > Check out my Work at: > LinkedIn: http://in.linkedin.com/in/thumsupdeicool > > "Contribute to the world, environment and more : http://www.gridrepublic.org > " > > > On Tue, Jul 27, 2010 at 7:47 PM, Jörg Godau <j.go...@schuetze-berlin.de>wrote: > >> Hi All, >> >> we have a fairly simple test that logs in to our application. >> >> We've setup a Gaussian Random Timer and are monitoring the results in an >> Aggregate report. >> >> My question is about the throughput - if we reduce the delays in the Timer, >> the time taken to log in to the application increases (which makes sense as >> there is more load on the server). >> >> Why is the throughput also increasing? If each request is taking much >> longer (ca. 10 times as long) when we increase the load - shouldn't the >> throughput be lower? >> >> Some numbers to illustrate: >> Timer ave 30 sec / deviation 15 sec => Average request 700ms, max >> 2760ms, throughput 3.1/sec >> Timer ave 10 sec / deviation 3 sec => Average request 9610ms, max >> 114300ms, throughput 5.0/sec >> >> Can someone please explain how this is possible? --------------------------------------------------------------------- To unsubscribe, e-mail: jmeter-user-unsubscr...@jakarta.apache.org For additional commands, e-mail: jmeter-user-h...@jakarta.apache.org
