Our network switch is nothing less than 1GB and the testing is happening internal to the network. So no chance of network contention here.
Secondly, after updating the JMeter version to 3.3 the CPU contention seems not to be appearing again. Also, as Antonio mentioned, the below mentioned bug fix sorted this. https://bz.apache.org/bugzilla/show_bug.cgi?id=59885 On Fri, Oct 20, 2017 at 9:42 PM, <[email protected]> wrote: > Are you sure it's just CPU? It may be bandwidth for the network device > that is constrained - and the CPU is trying to figure out where to put all > the packets. > > Keep in mind, images (especially) are a lot of bytes worth of data. If > your HTML is 10K, but you download a relatively small banner image - that > can be 100K of additional data (e.g., more than 10X the workload). Of > course that’s a lot more work for the system. A typical web page will have > 10-20KB of "skeleton" framework, but could have several megabytes of images > and non-HTML files. A multimedia page could have several times more of that > if you include streaming resources like movies or sounds. > > -- > Robin D. Wilson > Cell: 512-426-3929 > > > -----Original Message----- > From: Antonio Gomes Rodrigues [mailto:[email protected]] > Sent: Oct 20, 2017 2:39 A > To: JMeter Users List <[email protected]> > Cc: [email protected] > Subject: Re: Load Test with embedded resources chokes CPU > > Hi, > > Can you take a thread dump ? > > Antonio > > 2017-10-20 11:34 GMT+02:00 Bincy P S <[email protected]>: > > > We have changed the JVM heap size to 2G , Jmeter version to 3.3 and tried > > these test runs . I still see the same issue > > > > On Fri, Oct 20, 2017 at 2:56 PM, Philippe Mouawad < > > [email protected]> wrote: > > > > > What is the size of your JVM Heap ? If you didn't change it then it can > > > explain your issues. > > > Also why are you using an old version as 3.0, ? Last version is 3.3 ? > > > > > > > > > > > > On Fri, Oct 20, 2017 at 9:33 AM, Bincy P S <[email protected]> > > wrote: > > > > > > > Jmeter Configuration : > > > > Thread setting : 1 thread , 5 transaction controller ,each with 1 > HTTP > > > > sampler. Home Page, PLP,PDP ,AddtoCart,View Cart > > > > HTTP Request Sampler : tried with all Implementation,Download > embedded > > > > resources - retrieve embedded resources, parallel download - 6 > > > > > > > > No other settings were changed > > > > > > > > jmeter version : 3.0 > > > > > > > > On Fri, Oct 20, 2017 at 12:41 PM, Philippe Mouawad < > > > > [email protected]> wrote: > > > > > > > > > Hello, > > > > > What is your jmeter configuration ? > > > > > Did you change any setting ? > > > > > What is your jmeter version ? > > > > > > > > > > Thanks > > > > > Regards > > > > > > > > > > On Fri, Oct 20, 2017 at 9:06 AM, Bincy P S <[email protected]> > > > > wrote: > > > > > > > > > > > A load test with downloading just the HTML scales upto 300 users > on > > > one > > > > > > Load generator (8 Core, 8 GB RAM - CPU at 20% max and Memory at > > 5%). > > > > > > However, when a similar profile is tested with download of > non-HTML > > > > > > embedded resources (images, css, js, etc.,) along with HTML, the > LG > > > > gets > > > > > > utilized to the max (100% CPU, Memory still at 20%) with just 30 > > > users. > > > > > > > > > > > > Testing the same profiles with Load Runner, there is not much > > > > difference > > > > > on > > > > > > LG performance between the 2 profiles > > > > > > > > > > > > -- > > > > > > With Regards, > > > > > > > > > > > > Bincy Suresh > > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > Cordialement. > > > > > Philippe Mouawad. > > > > > > > > > > > > > > > > > > > > > -- > > > > With Regards, > > > > > > > > Bincy Suresh > > > > > > > > > > > > > > > > -- > > > Cordialement. > > > Philippe Mouawad. > > > > > > > > > > > -- > > With Regards, > > > > Bincy Suresh > > > > -- *Ganesh .N* *Bangalore* *Wireless:* *+91 9611906678*
