This is an interesting area. I would be fascinated to compare real with virtual hardware. We're deploying some fs hosts on "the cloud" and some of these hosted systems are pretty virtualised (eg. Linode uses Xen XPS) and some aren't (Hetzner being an example). It would be nice to start an area of the Wiki for this. It may be fairly random/adhoc contributions to start with, but I'm sure a structure could emerge with time. Any scripts used (with instructions) that could quickly be deployed to a host somewhere would be a good start. Also sound recordings of sound quality (together, perhaps, with measurements of latency) would help improve the authority of results. We would most probably be prepared to donate a day or so of testing. On 7 Jul 2012, at 09:05, Peter Olsson wrote:
Before trying anything else, use real hardware instead of virtual. A virtual server with this kind of load will only cause you trouble. Actually - most virtual solutions will cause you problems :)
Also, performance with average call duration of 4s is not really anything that would happen in real life.
But first of all - get real hardware.
/Peter
________________________________ Från: freeswitch-users-bounces@lists.freeswitch.org [freeswitch-users-bounces@lists.freeswitch.org] för Hanie Maghsoudy [h.maghsoudy@gmail.com] Skickat: den 7 juli 2012 09:46 Till: FreeSWITCH Users Help Ämne: [Freeswitch-users] Performance Testing
Hi all,
I searched for FreeSwitch call capacity, but most of the results wasn't new. So, I wanna ask if anybody has either tested FreeSwitch's performance recently, or got a dramatic result in real environment?
I tested call quality on this machine:
Virtual FreeSwitch server OS: CentOS release 6.2 - x86_64 CPU: 8 processor - Intel(R) Xeon(R) CPU X5670 @ 2.93GHz Memory: 8 G
After receiving incoming calls, FreeSwitch routed them to another sip server, without transcoding. The other server transmitted calls by playing an audio file. Meanwhile, I called an extension in FreeSwitch to test the call quality.
The result was like this:
1000 Concurrent calls Call duration: 160s Call rate: 6 cps (just creating channels) Max used Memory: 1416M Max CPU load: 0.24 Max Network throughput (recv/send): 6711k/80k Quality: Good
This test was taken before tearing down the channels.
Then, I took another test to estimate calls per second, and it wasn't what I was expected!
150 Concurrent calls Call duration: 4s Call rate: 30 cps (creating and tearing down) Max used Memory: 1293M Max CPU load: 4.50 Max Network throughput (recv/send): 828k/60k Quality: Average
And when I increase call rate to 50 cps:
1000 Concurrent calls Call duration: 4s Call rate: 50 cps (creating and tearing down) Max used Memory: 1730M Max CPU load: 29.9 Max Network throughput (recv/send): 1367k/202k Quality: Bad
Why call per second is such a big problem? Did anyone get a better result on this?
Thanks, Hanie
!DSPAM:4ff7e86232761844718139!
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
FreeSWITCH-powered IP PBX: The CudaTel Communication Server http://www.cudatel.com
Official FreeSWITCH Sites http://www.freeswitch.org http://wiki.freeswitch.org http://www.cluecon.com
Join Us At ClueCon - Aug 7-9, 2012
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org