jmeter-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Deepak Shetty <shet...@gmail.com>
Subject Re: Faster speed with higher threads
Date Mon, 22 Sep 2014 15:56:50 GMT
>I am not really sure on what basis to adjust up or down,
Ramp up is usually used to ensure that you dont get a burst of unrealistic
traffic.
 In your case you have indicated that a 1 user test concludes in 1.4
seconds. giving a ramp up of 100 seconds means that a new thread gets
created every second ,so say Thread 01 starts at time 0, Thread 1 starts at
time 1, but by this time Thread 0 has almost completed, Thread 2 starts at
time 2 but Thread 0 has already finished and has no impact on the server
anymore, so at any given point of time there might be probably about 3-4
threads active *only*- The server is usually impacted by concurrency and
your numbers are too low to have any impact on most servers i.e. whether
you have 10 or 100 users , the server sees only 3-4 or so threads at any
point in time so it should respond approximately the same

If on the other hand you really had 10 users concurrent slower than 100
users concurrent, then things that may make it respond faster
a. Are you using unique users? If same (i.e. you dont have 100 distinct
users then you might see improvements )
b. What data are they accessing (there are various types of shared data
like say menus ) - if the data being accessed is shared and cached you
might see it. Webservers cache data too for things like static files
c. Certain application servers like hotspot JVMs show better responses with
more requests .
d. You didnt use assertions on your tests - what you are actually seeing is
quick returning error pages :)




On Mon, Sep 22, 2014 at 7:47 AM, rxfillpharm <rxfillpharm@yahoo.com.invalid>
wrote:

> Yes, threads=users.  100 Users=100 Threads.
>
> Jmeter user pages http://jmeter.apache.org/usermanual/test_plan.html
> suggest
>
> "Start with Ramp-up = number of threads and adjust up or down as needed."
>
> I am not really sure on what basis to adjust up or down, so I just
> leave ramp equal to number of threads (aka users).
>
> But does anyone know why, on average, per thread speed is faster when
> there are more threads/users.
>
>  1 thread    = 1.402s
> 10 threads =1.707s
> 50 threads =1.087s
> 100 threads =0.813s
>
> I thought the more threads, the busier the servers and the slower the
> speed.
>
> --------------------------------------------
> On Fri, 9/19/14, Deepak Shetty <shettyd@gmail.com> wrote:
>
>  Subject: Re: Faster speed with higher threads
>  To: "JMeter Users List" <user@jmeter.apache.org>
>  Date: Friday, September 19, 2014, 4:43 PM
>
>  Im assuming that for you
>  1 thread = 1 user - so when you say 100 users u
>  actually mean 100 threads?
>  >Ramp is equal to the number of threads.
>  so you have set ramp up to 100 seconds for 100
>  users? You aren't generating
>  much
>  concurrent load if one test is usually done in 1 -2
>  seconds.
>
>
>
>
>  On Fri, Sep 19, 2014 at
>  12:59 PM, rxfillpharm <rxfillpharm@yahoo.com.invalid
>  > wrote:
>
>  > Hi,
>  > My test signs on
>  and signsoff a web site. Ramp is equal to the number of
>  > threads.
>  > Here are the
>  results:
>  >
>  > (First we
>  recycle all servers)
>  > 1 user= 1.402s
>  > 10 users=1.707s
>  > 50
>  user=1.087s
>  > 100 users=0.813s
>  >
>  > Why are 100 users
>  faster on average, than fewer users?
>  >
>  > My education in jmeter is expanding but I
>  still have the nagging questions
>  > .
>  > 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