Another thing to consider is how much network bandwidth you are using. If each request is returning 100K in data, your 25 rps rate is ~2.5MB per second (which is ~25 M bits per second). Depending on your network interface and infrastructure that could be a fraction of the available bandwidth or all of it. Keep in mind, your are limited by the slowest hop in the network.
For example, we test on a separated and isolated network switch that has 2Gbps of total throughput. But we have several machines on that switch (all part of our test) and they have to share the 2Gbps between themselves. Moreover, each machine has only a 1Gbps NIC. That is generally more than enough for our testing, so we aren't usually constrained by network capacity. However, If we tried to run this same test to our production system, even though those servers have much higher capacities than our test environment, our internal network has only our 25Mbps ISP connection to the Internet (which is shared by the whole company), and we'd be throttled by that connection to our production servers (which are in a hosting facility in a different city). So even though our test environment is not nearly as beefy as our production setup, our tests to the production systems would be much slower because of the limited network capacity. Also, the fact that the Internet connection is shared with people who are not part of the test means that we would get widely varying results depending on what other people are doing when we fire up a test (for example, downloading updates to Windows or streaming movies/music). But in our test environment we share the network only with our own test servers, so we can control that variable during testing. Another thing to check is how much your load generator is recording during the tests. We have a tree listener in our test plans, but we found that if we record "all" responses, the our Jmeter load generation is severely reduced. Moreover, it degrades even further the longer the test runs. However, if we use the tree listener to only record "errors" instead of everything, it works without interfering in the test, and we still get enough information to figure out what broke when something goes wrong. (Although I'd still like to see a feature to record the whole thread when any iteration of the thread has a failure.) The best thing to remember is that you are testing a "system", and it is constrained by the most limited part during your test. Figuring out which part is the bottleneck is all the fun... -- Robin D. Wilson VOICE: 512-777-1861 On May 23, 2013, at 4:34 AM, Adrian Speteanu <asp.ad...@gmail.com> wrote: Hi, Not being able to achieve target throughput is not always related to things that are wrong on the test client (considering that 25 rps seems rather low ), but rather with applications deployed in a state that is not ready to handle the generated traffic. The advise below is very good, so that you eliminate common/obvious mistakes client-side. But I also suggest you keep an eye for possible application & setup specific problems. Cheers, Adrian S On Thu, May 23, 2013 at 12:30 AM, Shmuel Krakower <shmul...@gmail.com>wrote: > Things you may try: > 1. Remove the Throughput controller and see what happens. > 2. Add more threads and see what happens. > 3. Is your load machine is overkilled (CPU at 100%, etc...)? > 4. If all above leaves you with same limit of about 25MPS it might be the > target service which cannot handle anymore load. Try: > a. Running at the same time, the same load from another machine and see if > you get to 50MPS in total or still you get about 25... > b. Look into the service under load and try to figure out what is the > bottleneck > c. Maybe something else between the two machines which runs JMeter and the > Service is limiting you? (i.e. internet connection limits) > > > > Shmuel Krakower. > www.Beatsoo.org - re-use your jmeter scripts for application performance > monitoring from worldwide locations for free. > > > On Wed, May 22, 2013 at 6:48 PM, Sameera Rao P <samee...@techmahindra.com >> wrote: > >> Hi >> >> I Am using soap sample xml / RPC and trying to post ebxml messages. I am >> not able to achieve my load of 100 Messages Per second running from GUI. >> I tried using constant throughput timer with 15 threads ,but still not >> able to go beyond 25MPS. Can you please suggest the solution for >> achieving the load >> >> Thanks & Regards >> Sam >> >> >> > ============================================================================================================================Disclaimer: >> This message and the information contained herein is proprietary and >> confidential and subject to the Tech Mahindra policy statement, you may >> review the policy at <a href=" > http://www.techmahindra.com/Disclaimer.html >> ">http://www.techmahindra.com/Disclaimer.html</a> externally and <a > href=" >> http://tim.techmahindra.com/tim/disclaimer.html";> >> http://tim.techmahindra.com/tim/disclaimer.html</a> internally within >> Tech >> > Mahindra.============================================================================================================================ >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: user-unsubscr...@jmeter.apache.org >> For additional commands, e-mail: user-h...@jmeter.apache.org >> >> > --------------------------------------------------------------------- To unsubscribe, e-mail: user-unsubscr...@jmeter.apache.org For additional commands, e-mail: user-h...@jmeter.apache.org
