Hi First When you need to limit the number of requests /per second use a constant throughput timer.
The explanation for the behavior you see is If you dont have any timers or anything but just a ramp-up of say 1 second and say your pixel takes 0.1 seconds then Start of test , JMeter will create 1 thread , thenmake a request , get the response back in 0.1 seconds and immediate go and make the second request , and so on for the first second , then JMeter will created the second thread(ramp up of 1 sec) so that now every 0.1 seconds two threads will request the pixel and at the third second Jmeter creates thread 3 and so on till the 100th second by which time all 100 threads are created after which ramp up does not apply and all of them make a request every 0.1 second. So you will see a much higher throughput (it will be determined by how quickly the response returns) In the second case , with constant timer , it might seem to work for you because your pixel probably takes very little time. Consider however if your pixel request took 0.5 seconds to process. Then a thread would make a request get the response back in 0.5 seconds and then wait the entire 1 second before making the next request - so you would be making 1 request per 1.5 second per thread which is lower than what you want. The Constant throughput timer tries to account for the above so if you say 1 want 60 requests per minute , per thread then it will adjust its delay accordingly (assuming that your system can support that throughput) regards deepak regards deepak On Thu, Jul 1, 2010 at 6:51 AM, nandu2008 <[email protected]> wrote: > > Hey Deepak, > > Sorry I didn't get it completely. > Here is the case: > 100 threads per second and forever(I quit after a certain amount of time, > say 30 mins). > So, I should be seeing 100X30X60 results in the response if it had run once > per second...right?(what I am testing is a pixel. So, response is very > fast). > But, I see more than that amount. Almost triple of that amount.(I > verirified > that all has a successful reponse code-200). > SO, I am assuming that each thread ran more than once per second. > Is that RIGHT? > > Then, if I give a constant timer(1000ms), I see the exact number of results > in the result set. 100X30X60. > So, that means that, in my case, the threads ran only once per > second...right? > So, in this case, giving a constant timer is okay...right? since I saw the > exact number of results. > > Or am I missing something somewhere? > thanks and sorry Deepak to bug you again, > N > > > Deepak Shetty wrote: > > > > Hi > > Use a throughput controller to control how many requests per second are > > sent. > > > > JMeter will send as many requests as it can. The rampup merely means how > > long Jmeter will wait before creating a thread at the start. This has no > > meaning after all the threads are created . A delay only guarantees a > > delay > > between requests - it does not factor in the time taken to actually > > execute > > the request so your throughput will be lower than what you want > > > > regards > > deepak > > > > On Wed, Jun 30, 2010 at 9:42 PM, nandu2008 <[email protected]> > wrote: > > > >> > >> Hey Deepak, > >> thanks again. > >> got it right now. > >> > >> one more question. > >> I am trying to load an application with 100rps for 30 minutes. > >> So, I am doing this by taking 100 different threads and running it for > >> ever > >> and stopping the test after 30 minutes. > >> I am giving a ramp up of 1 sec. > >> In the result data, I see that there are more requests than what I > should > >> be > >> getting. > >> I need to get 100X30X60 total in 30 minutes. > >> But the result is much more than that. > >> I am assuming that each thread is running more than once per second. > >> IS my assumption RIGHT?otherwise, I would have got 100X30X60 rps in > >> result > >> as well. > >> > >> So, in order for each thread to run only once per second, I am giving a > >> constant timer of 1000. > >> Is that right? > >> Does it actually mean that each of these 100 threads will run only once > >> per > >> scond? > >> Is this the right way to make sure that only 100 threads will be run > >> ineach > >> second? > >> > >> thanks, > >> N > >> > >> > >> Deepak Shetty wrote: > >> > > >> > >see that there is / appended at the end of the pixel.(jmeter). > >> > Im not sure what you mean. Do you mean the URL has a forward /? If > >> your > >> > original request doesnt have this , it is your web server which > >> redirects > >> > you to a URL with a '/' (because your request was for a directory) - > >> after > >> > which the default content for the directory is probably served . > >> > use follow redirects and view results tree listener to see > >> > > >> > regards > >> > deepak > >> > > >> > > >> > On Tue, Jun 29, 2010 at 4:45 PM, nandu2008 <[email protected]> > >> wrote: > >> > > >> >> > >> >> Thank you Deepak. > >> >> > >> >> Can I ask you one more q? > >> >> I see that there is / appended at the end of the pixel.(jmeter). > >> >> But, if I try it in a browser, I see that there is no / at the end. > >> >> Actually, we need to get the pixel without the / in jmeter. > >> >> > >> >> ps: I see that if I try www.google.com, there is a / at the end in > >> both > >> >> jmeter and in browser window. > >> >> > >> >> thanks, > >> >> Bindu > >> >> > >> >> > >> >> > >> >> > >> >> nandu2008 wrote: > >> >> > > >> >> > Hey Deepak, > >> >> > > >> >> > Can I ask you one more q. > >> >> > So, I see the load time as expected, but again the latency is zero. > >> >> > So, I am assuming that the result is fine. > >> >> > Do you know whether the http_request/http client sampler will > record > >> >> > latency?(Might be that it doesn't record latency). It is fine in > >> >> > > >> >> > Also, one more q. > >> >> > I am sending request for an http image pixel. > >> >> > But, I see that the response header field in sampler result(in > >> listener > >> >> > "view tree") > >> >> > is http. But, the request field has the https url I mentioned. Does > >> it > >> >> > mean that my test didn't run as it should be? > >> >> > > >> >> > It looks as follows: > >> >> > > >> >> > Response headers: > >> >> > HTTP/1.0 200 OK > >> >> > Content-Type: application/octet-stream > >> >> > Content-Length: 42 > >> >> > Allow: GET > >> >> > > >> >> > thanks again, > >> >> > Bindu > >> >> > > >> >> > > >> >> > nandu2008 wrote: > >> >> >> > >> >> >> hello all, > >> >> >> > >> >> >> I am testing an https image pixel. > >> >> >> So, I don't have any authentication like user name etc. > >> >> >> > >> >> >> Basically, I run this test on a mac amchine and teh result is > good. > >> >> >> But, if I run it an a linux machine, I see that the latency is > >> always > >> >> 0. > >> >> >> also, I see some INFO in the logs as follows: > >> >> >> > >> >> >> INFO - org.apache.commons.httpclient.HttpMethodDirector: I/O > >> exception > >> >> >> (java.net.SocketException) caught when processing request: > >> Connection > >> >> >> reset > >> >> >> INFO - org.apache.commons.httpclient.HttpMethodDirector: Retrying > >> >> >> request" > >> >> >> > >> >> >> What does that mean? > >> >> >> is that an error? is that the reason why i am not getting latency? > >> >> >> > >> >> >> any help is appreciated. > >> >> >> > >> >> >> thanks, > >> >> >> Bindu > >> >> >> > >> >> > > >> >> > > >> >> > >> >> -- > >> >> View this message in context: > >> >> > >> http://old.nabble.com/jmeter---https--question-tp28997028p29029543.html > >> >> Sent from the JMeter - User mailing list archive at Nabble.com. > >> >> > >> >> > >> >> --------------------------------------------------------------------- > >> >> To unsubscribe, e-mail: [email protected] > >> >> For additional commands, e-mail: [email protected] > >> >> > >> >> > >> > > >> > > >> > >> -- > >> View this message in context: > >> http://old.nabble.com/jmeter---https--question-tp28997028p29035934.html > >> Sent from the JMeter - User mailing list archive at Nabble.com. > >> > >> > >> --------------------------------------------------------------------- > >> To unsubscribe, e-mail: [email protected] > >> For additional commands, e-mail: [email protected] > >> > >> > > > > > > -- > View this message in context: > http://old.nabble.com/jmeter---https--question-tp28997028p29045878.html > Sent from the JMeter - User mailing list archive at Nabble.com. > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
