[Serusers] Test Tools for SER

Andrei Pelinescu-Onciul pelinescu-onciul at fokus.fraunhofer.de
Thu Jun 10 15:16:46 CEST 2004


On Jun 09, 2004 at 15:12, Ezequiel Colombo <ecolombo at arcotel.net> wrote:
> Hi, i compiled some of the tools included in the /test directory of the source CVS of SER. I want test the performace of SER in my server and look at the udp_flood.c program included, this generate a flood of SIP messages to the server. My question is how interpret the results.
> 
> If i send for example a flood with INVITE request for non-existent user to my SER and see how much SER respond to this flood in a time period, i supose that this is the Call Per Second supported by SER ?

No. In our test we understand by call the following messages:
1. INVITE remote_party
2. 100 Trying
3. 180 Ringing
4. 200 Ok
5. ACK
6. BYE
7. 200 Ok
8. ACK

Also the speed depends on whether you are using stateless (faster) or
statefull forwarding and in general the contents of your cfg script.
> 
> Ex:
> 
> # flood SER with 100 request in one second
> ./udp_flood -f invite00.sip -d 192.168.1.1 -p 5060 -s 1000000 -t 100 -v -c 100
> 
> # SER make a log(1,"PROC_OK") for each invite processed
> ...
> 
> # At the end of udp_flood execution + some seconds i make
> grep "PROC_OK" /var/log/ser.log | wc -l
> 
> Now, after this if i can see 100 at result of wc -l i supose at least 100 INVITE per second is supported for my SER ???

We tested at some point ser on an iPaq pda (intel strongarm 110 cpu) and
it did over 100 cps in statefull mode (IIRC arround 150).
So I assume your PC can do much more than 100cps (unless you don't exec
 from the script some very time consuming application).

A dual Athlon MP2000+ does 4800-4900cps and probably arround 6000-7000
in stateless mode.


Andrei




More information about the sr-users mailing list