Hello, Thanks for the advice, it's much appreciated.
One more question. You mentioned recording actual requests issued by the browser and replaying them. Looking through the documentation I see this in the wiki under 'Recording Tests': http://jmeter.apache.org/usermanual/jmeter_proxy_step_by_step.pdf Is this what you were referring to? Thanks, EJ Nivara On Tue, Jul 22, 2014 at 3:42 AM, Ryabtsev Vladimir <[email protected]> wrote: > Hi, > > You can quickly switch test for using and not using proxy in HTTP Request > Defaults element. > I'd disable cache manager for your case as you must compare response times > with and without proxy for absolutely identical scenarios. With cache > manager number of actual requests is going to be bit unpredictable. > > You should understand some things about 'Retrieve All Embedded Resources' > option: > 1. You will see summary time for you HTTP Request sampler in logs and > reports. This time is sum of time of request for '/' page and times of all > requests for embedded resources. (You can investigate actual requests and > their times in sub-samplers in tools like View Results Tree, but this is > only for debugging purposes, disable VRT before real test.) At the same > time you may be interested in times for requested resources because this is > important for proxy (it may cache resources, but it can't cache dynamic > pages). > 2. This option will not provide issuing requests for dynamically loaded > resources (e.g. via JavaScript, Flash and so on). That is, you should not > expect full equivalence with browser requests. > > This means you should decide whether it suits you, basing on aspects you > want to test. You might want to record actual requests issued by browser > and replay them. In this case you'll get separate sampler for each resource > and you'll be able to get their times in reports. > > Concerning listeners, try Aggregate Report for essential statistical > information (min, max, avgs). > > Also, 20 iteration is not enough for test, consider running long tests (1 > hour for example). Take into consideration current load of web site, it > should be the same in both tests with and w/o proxy. > > ----- > VR > > -----Original Message----- > From: EJ Nivara [mailto:[email protected]] > Sent: Monday, July 21, 2014 9:52 PM > To: [email protected] > Subject: Advice for approximating browser behavior > > Hello all, > > I'm in the process of setting up benchmark testing to various web sites > with and without our proxy server. I've set up JMeter and wanted some > advice on whether I am approximating generic out-of-the-box browser > behaviour appropriately. Essentially we are trying to compare the > differences in latency between certain sites with and without our proxy. > > So far my current set up is as follows: > > 1 Thread Group (nothing special, 1 user, 1 second ramp-up period, 20 > iterations > > The thread group has the following config elements: > > 1 HTTP Request Defaults (Retrieve All Embedded Resources checked, as it > seems it'd be useless without retrieving all of the objects a normal > browser would) > 1 HTTP Cache Manager (defaults) > 1 HTTP Cookie Manager (defaults) > > Samplers: > > 1 HTTP Request (/ for path as I am just testing the index page - for now, > will modify or add additional pages as required, Retrieve All Embedded > Resources, again as this seems the right thing to do) > > Listeners: > > Right now I'm just using 'Graph Results' as they want visual, but I > wouldn't mind advice for collecting other relevant data. > > I'd appreciate any advice, as this is somewhat new to me. > > Regards, > > EJ Nivara > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] >
