> By the way, you don't say how you measured the actual times - were you > aware that the timestamps in JTL files are END times, rather
Sorry Seb, this is how I measured the actual times. I didn't use the JTL files (at my settings they only record the sample results, not delays). I measured the times using a wrapper script to invoke JMeter. First, I set delay on all timers to 0 and did a few test runs. The average - 12 seconds - gave me a base line time for the test with no delay. I then increased the delay in all 5 timers in 500 ms increments (by modifying a runtime property). I ended up with 23, 37, 49, 62 seconds for the last four results in the table below. I then subtracted 12 seconds from each run, then divided it by 5 (there were 5 timers in serial) to get the figures below: > > Constant delay Approx. actual delay. > > (ms) (ms) > > ======================================= > > 0 0 > > 500 2000 > > 1000 5000 > > 1500 7500 > > 2000 10000 > > ======================================= If it is a bug, it is really hard to reproduce. I saw this effect in another simple JMX, but only if I inserted a HTTP sampler between timers (the sampler itself loads a lightning fast local page, with 0 page load time). When the timers had nothing between them, they seem to delay correctly. With regards, Sonam Chauhan -- Corporate Express Australia Ltd. Phone: +61-2-9335-0725, Fax: 9335-0753, Email: [EMAIL PROTECTED] > -----Original Message----- > From: Sebastian Bazley [mailto:[EMAIL PROTECTED] > Sent: Wednesday, 7 January 2004 12:08 PM > To: JMeter Users List > Subject: Re: Timer wierdness > > This could perhaps be due to other activity on the system. > Or it could be a bug.... > > By the way, you don't say how you measured the actual times - were you > aware that the timestamps in JTL files are END times, rather > than start times ? > (the latest code in CVS can save start times instead, if the appropriate > property is defined before starting JMeter) > > S. > ----- Original Message ----- > From: "Sonam Chauhan" <[EMAIL PROTECTED]> > To: "'JMeter Users List'" <[EMAIL PROTECTED]> > Sent: Wednesday, January 07, 2004 12:19 AM > Subject: Timer wierdness > > > > Has anyone seen timers taking longer than they should? I have a JMX that > > makes 5 HTTP requests with Uniform Random Timer inserting a delay in > between > > each request. > > > > Setting the timer's random delay to 0, and varying the timer's constant > > delay, I got the following figures for the real delay of each timer. > > (numbers adjusted for page load delays.) > > > > Constant delay Approx. actual delay. > > (ms) (ms) > > ======================================= > > 0 0 > > 500 2000 > > 1000 5000 > > 1500 7500 > > 2000 10000 > > ======================================= > > > > In this case, the timer actually seem to take 4-5 times as long. > > > > With regards, > > Sonam Chauhan > > -- > > Corporate Express Australia Ltd. > > Phone: +61-2-9335-0725, Fax: 9335-0753, Email: [EMAIL PROTECTED] > > <mailto:[EMAIL PROTECTED]> > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
