How to set constant request per second using throughput shaper timer

564 views
Skip to first unread message

swetha...@gmail.com

unread,
Dec 17, 2012, 1:34:27 AM12/17/12
to jmeter-...@googlegroups.com
Hi People,

My requirement is to run performance test for 10min at request rate 250 RPS.
I am setting start and end RPS in Throughput shaper timer as 250 for 600 sec duration.
Also I am using Thread Group of 23 threads with zero ramp up time.
When I start the test I am getting Transaction  per second as expected to be 250 RPS but after 4min of test run, RPS is getting dropped eventually to 120RPS.
I tried by lowering the number of threads in Thread Group but it didn't help.
Please let me know if anybody has solution for it and whether dropping in Transaction per second is related to server capacity to handle requests.


-Thanks
Swetha

Andrey Pohilko

unread,
Dec 17, 2012, 5:48:58 AM12/17/12
to jmeter-...@googlegroups.com
The most common reason for this is that your threads pool is not enough to support your RPS. THere may be several reasons for this:
a) You have enough threads at the start, but they have limited duration or loops, and some threads are finished after 4 minutes. Please verify that threads count not decreases with Active Threads Over Time graph.
b) Your service have increasing response time, and after 4 minutes that time is so high, and threads pool of 23 is not enough anymore. Verify this with Response times over time graph.

To fight scenario 'a' you should configure your threads not to stop on error, and have infinite loops.
To fight scenario 'b' you should increase threads pool, there can be no other solution - JMeter needs active threads to generate request.

понедельник, 17 декабря 2012 г., 10:34:27 UTC+4 пользователь Swetha Narayanappa написал:

Swetha Narayanappa

unread,
Dec 17, 2012, 2:14:36 PM12/17/12
to jmeter-...@googlegroups.com
Thanks Andrey for your quick response.
I looked into Active Threads Over time graph and there is no drop in the thread counts.
I also checked response times over time graph and there is no issue.
I am suspecting that my client machine(linux box) is not able to send so many request and planning to run from hudson machine.


-Thanks
Swetha
Reply all
Reply to author
Forward
0 new messages