Thank you to both of you, (Deepak Shetty/John), It is very clear now, I never ran for several iterations but i was trying to load concurrent users reducing rampup time, which is wrong. il consider all the facts and try again ,thanks
On Fri, Feb 28, 2014 at 3:54 AM, Deepak Shetty <[email protected]> wrote: > Hi > a) Ramp up time is a testing tool concept usually used so that your server > doesnt see a burst of traffic that is unrealistic - For e,g, if say you > have 100 users "active" but only 10 concurrent you might have modeled this > as a thread group with 100 threads - however when you started up your test > with 0 rampup time it might be that you saw 100 requests concurrently on > the server - which is unrealistic for the example - so you give a rampup > time so that the initial requests are spaced out (it helps your client > jmeter too!) > > Think time on the other hand is the time between requests (e.g. the time > spent by a user reading a page or filling the form) - whether you use think > time or not is dependent on what is the objective of running the test , > what reports are expected and what your application behavior is like. If Im > trying to simulate race questions , I will rarely,if ever, have think times > in my tests. > > > i couldn't load all 100 users threads concurrently > Did you run it for multiple iterations ? otherwise the thread will die as > soon as its tasks are over and if you have a rampup time that is higher > then the time a thread needs for the test then obviously you wont be able > to simulate 100 threads. You also usually need more threads on the client > than the concurrent usage on the server(dependent on script) and you might > need more than one JMeter instance (depending on many factors) > > > > On Tue, Feb 25, 2014 at 11:46 PM, Perf Test <[email protected]> wrote: > > > Dear Deepak Shetty, > > > > I'm so glad to see your detailed clarification, I think you are right, > > However if i'm going to do a concurrent users load test, according to > your > > case i will have to chose option b) . then shouldn't we require to use > > think time with option b)? > > > > When i use think time with option 'b)' i couldn't load all 100 users > > threads concurrently (but all the 100 logins successful) , but when i > > remove the think time i.e. reduce the rampup time i was able to load 100 > > concurrent users. do you think it is right way of testing concurrent > users > > load? > > > > Thanks, > > Sam > > > > > > > > > > On Tue, Feb 25, 2014 at 10:40 PM, Deepak Shetty <[email protected]> > wrote: > > > > > No it depends on your definition of "active" - the common definitions I > > > have heard of is > > > a) active - has a session on your system but not necessarily doing > > anything > > > - may be reading the article or typing in a form but not clicking > submit > > ( > > > b) concurrent - actually doing something like clicked a link or > > submitted a > > > form > > > (though I have also heard it being used interchangeably) > > > For these definitions > > > a) only has meaning for memory usage and if you dont use the session > > much , > > > its not really relevant except when you want to closely model real > world > > > behavior (think times etc) and check cache performances and expiry or > if > > > you do use session then it is important > > > > > > b) is usually the more relevant attribute for performance tests. > > > > > > > > > > > > On Tue, Feb 25, 2014 at 4:08 AM, Deepak Goel <[email protected]> > wrote: > > > > > > > It is probably the same > > > > On 25 Feb 2014 11:42, "Perf Test" <[email protected]> wrote: > > > > > > > > > Dear All, > > > > > > > > > > What is the Difference between JM 'concurrent users' load of 100 > and > > > > > 'Active users' 100 testing for a given time period. > > > > > > > > > > > > > > > Thanks. > > > > > Sam > > > > > > > > > > > > > > >
