[OpenSIPS-Users] opensips performance testing

[Tarik Tankovic]

Hi Vlad,

I am using the following hardware:

CPU: Intel  CORE i5-2400 3,1 GHz
RAM: DDR III 4 GB 1333 MHz
network: 100 Mbit/s ethernet

and OS is debian linux (see attach).

I think the conditions were more or less the same ones.

I have two virtual machines with debian linux installed on this
configuration of mine and I have dedicated 512 MB of RAM memory to
each of them. These two instances of SIPP are running on these vritual
machines and the OPENSIPS is on the main machine.

I have just used Debian system monitor tool both on the main machine
and also on the two virtual machines and here are the results:

Main machine:
CPUs: idle state: 15%-20%, test with 1 child 40%, test with 4 and 10
children 45%
memory: idle state: 510 MB, test with 1 child 515 MB, test with 4
children 522 MB and test with 10 children 529 MB (max. 3,9 GB)
network: idle state: 0, all the tests max. 1,4 Mbit/s

virtual machines:
CPU: idle state: 25%, test with 1 child 60%, test with 4 children 65%,
test with 10 children 70%
memory: idle state: 123 MB, test with 1 child 129 MB, test with 4
children 130,2 MB and test with 10 children 130,8 MB (max. 502 MB)
network: idle state: 0, all the tests max. 650 kbit/s

So there is obvious increase in components usage but I can't see that
even one of these is going into the saturation or 100% usage.

Thanks for helping me Vlad!

Greetings,

Tarik

On 1/23/12, Vlad Paiu <vladpaiu at opensips.org> wrote:
> Hello,
>
> That is indeed strange. What machine were you running on, in terms of
> hardware specs ?
> Also, are you sure the runtime conditions ( CPU load, network load ) for
> all the tests was the same ?
> Can you check what is the actual bottleneck when using either 1,4 or 10
> children, be it CPU or bandwidth, etc ?
>
> Regards,
>
> Vlad Paiu
> OpenSIPS Developer
>
>
> On 01/21/2012 09:17 PM, Tarik Tankovic wrote:
>> Hi,
>>
>> I am doing some OPENSIPS performance testing by using two SIPP's. One
>> SIPP is using a little bit modified UAC scenario and the second one is
>> using a little bit modified UAS scenario. OPENSIPS is acting as a
>> stateful UDP proxy without authentification.
>>
>> I have changed only one parameter in this test bed of mine and that is
>> children parameter in Global parameters section of the opensips.cfg
>> file. I was a little surprised to see that the best result
>> (throughput, presented in the form of cps) I got when used children=1,
>> than children=10 and after that children=4.
>> I did 5 tests for each value (of the children parameter) that lasted
>> 180 seconds and after that I did some averaging.
>>
>> Can anyone give me some explanation for this strange result of mine?
>>
>> Thanks,
>>
>> Tarik
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opensips.org
>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>
> _______________________________________________
> Users mailing list
> Users at lists.opensips.org
> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Screenshot-System Monitor.png
Type: image/png
Size: 48615 bytes
Desc: not available
URL: <http://lists.opensips.org/pipermail/users/attachments/20120123/85b21239/attachment-0001.png>