Throughput Shaping Timer does not work accurately with small rps

[matthias...@googlemail.com]

Hello,

I want to run a load test, but the ThroughputShapingTimer won't run as it is supposed to.

I need to place a SOAP/RPC-Test every 2 seconds.  Of course you will say this is a constant throughput Timing, but in the next steps I need to shape the load.

So I filled the plugin with 0.5 at start and end and I set 30 Threads (without repeating) because I dont want the final peak, which would kill my target system :o)

Just have a look at the start times and see, that this is not one request every second second.

17:55:36

17:55:38

17:55:42

17:55:46

17:55:50

17:55:54

17:55:58

17:56:02

17:56:06

17:56:10

17:56:14

17:56:18

17:56:22

17:56:26

17:56:30

17:56:34

17:56:36

17:56:40

There is a delay in the start-times of the probes, and many have a difference of something about 4 seconds. I don't know why.

Have a look, what happens, if I set start an end rate to 1.0 requests per second.
Should I expect that there will be a request generated every second?

19:31:03

19:31:04

19:31:06

19:31:08

19:31:10

19:31:12

19:31:14

19:31:16

19:31:18

19:31:20

19:31:22

19:31:24

19:31:26

19:31:28

19:31:30

19:31:32

19:31:33

19:31:35

19:31:37

19:31:39

19:31:41

19:31:43

19:31:45

19:31:46

19:31:48

19:31:50

19:31:52

19:31:53

19:31:55

19:31:56

Fo now, I'll get nearly every second second a request, what I was intended to at the beginning. But it runs unregularily and not the way it shoult to.  It skips seconds. Is this an error in the plugin or a setup error?

[Andrey Pohilko]

Hi,

1. How much threads do you have in your setup? 

2. And also how long does your service respond?

3. Do you see any messages related to the shaper in jmeter.log?

--

Andrey

среда, 16 апреля 2014 г., 21:39:12 UTC+4 пользователь matthias...@googlemail.com написал:

[matthias...@googlemail.com]

Hi,

1. Threads Count is 30 for this is just the test of a bigger one
Ramp-Up Period is 0
Repeat Once
Settings in Throuput Shaping Timer are for a short period of 60 seconds.

2.Some Values

Duration	Latency
5338	5257
7254	7162
4852	4770
4125	4021
3994	3918
4285	3839 3. there are no related messages in jmeter.log

[Andrey Pohilko]

It looks almost ok. 

I must admin that the shaper is doomed to work not very well for small request rate. It's just because of its "timer" nature. The response times are quite big and this adds some difficulty to the shaper.

I have no idea how to mitigate this effect. Can you try starting more threads and set their loop to infinite?

среда, 16 апреля 2014 г., 23:33:22 UTC+4 пользователь matthias...@googlemail.com написал:

[matthias...@googlemail.com]

Hi, I did this at the beginning.
But the effect is the same. I decided to do it this way because you know the problem with the final spike.

I wanted to consider all Requests till the timeout at 60 Seconds. With the expectet 0.5 Requests per second this would be 30 Threads.

30 Threads at the end of the tests as final spike was not recommendable for the target system. And however, the timing problem continued to exist.