jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Deepak Shetty <shet...@gmail.com>
Subject Re: JMETER performance - Processes & threads
Date Wed, 29 Oct 2014 21:52:58 GMT
Hi
Typically when you are getting different results when you change the way
Jmeter is hitting the server(as in your case) without actually changing
what the server sees (in your case it should be seeing ~100 requests in
parallel) then there is some issue on the client that needs to be looked
at. Server issues will be after you fix your client.

I suppose you are looking at Xmx for JVM client which you say is 4GB - How
much is total on linux (free -m)
When you run 2JVMs you are making double the memory available , which for a
GC dependent system like Java should be significantly faster.
But 30 seconds looks to be too large for your scenarios(unless u have a
really slow app) hence you need to see if your JMeter is telling you 1 page
= 30 seconds but a browser at the same time is telling you for e.g. 5
seconds for the same page , then likely you are running too many threads
for your client machine or your network is choked or something. if the
browser is telling you something close to 30 seconds then you are good.



On Wed, Oct 29, 2014 at 2:08 PM, rxfillpharm <rxfillpharm@yahoo.com.invalid>
wrote:

> Thank you Deepak.
>
> Should I be concerned about memory on the jmeter client, or on the server?
>
> Regarding the system handling that much load, where is the issue-- on the
> machine
> jmeter is running or on my application server?
> I am running the exact same test, just changing the number of threads:
>
> TRIAL ONE:
>    one jmeter session running 100 threads by 1 run.   100 seconds
>
> TRIAL TWO:
>    two jmeter sessions (different unix windows) running 50 threads by 1
> run each.    30 seconds each
>
>
>
> --------------------------------------------
> On Wed, 10/29/14, Deepak Shetty <shettyd@gmail.com> wrote:
>
>  Subject: Re: JMETER performance - Processes & threads
>  To: "JMeter Users List" <user@jmeter.apache.org>
>  Date: Wednesday, October 29, 2014, 12:42 PM
>
>  most likely memory - If
>  your OS has enough spare memory and you start up 1
>  JVM with for e.g. 1GB RAM heap , it will be
>  slower than if you start 2 JVM
>  with the same
>  memory (so effectively 1*2 =2GB RAM).
>  You
>  might get the same result by simply increasing your memory
>  when you
>  start up your JVM (the Xms and Xmx
>  parameters). - Just a guess - only you
>  have
>  access to your machine
>
>  But
>  its also likely that your system isnt actually able to
>  generate that
>  much load if you are seeing
>  the swing you are seeing - What is the response
>  time as seen by your browser on a totally
>  different machine while your
>  Jmeter is
>  showing 30 seconds average? if its much lower (note
>  compare
>  apples to apples i.e. look at the
>  indvidual requests using something like
>  firebug and network tab rather than page render
>  times) then you have an
>  issue
>
>
>
>  On
>  Wed, Oct 29, 2014 at 8:50 AM, rxfillpharm <rxfillpharm@yahoo.com.invalid>
>  wrote:
>
>  >
>  Hi,
>  > I hope someone can clarify a jmeter
>  performance question.
>  >
>  > I have a test that simulates user
>  sign-ins.
>  >
>  > When I
>  run jmeter in a unix process with 100 threads, the average
>  speed is
>  > 100sec.
>  >
>  > When I run 2 separate
>  unix processes with 50 threads each, starting each
>  > process within
>  >  1
>  second of each other,  the average speed is @ 30 secs.
>  >
>  > What is going on
>  here?  What is the bottleneck in the 100 thread process
>  > that slows things down, and
>  > is there any way to speed it up?
>  >
>  > Thank you.
>  >
>  >
>  >
>  >
>  ---------------------------------------------------------------------
>  > To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
>  > For additional commands, e-mail: user-help@jmeter.apache.org
>  >
>  >
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: user-unsubscribe@jmeter.apache.org
> For additional commands, e-mail: user-help@jmeter.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message