[Freeswitch-users] Network Card

Steven Ayre steveayre at gmail.com
Fri Aug 15 11:37:03 MSD 2014


Have you tuned the core and sofia databases - moving it onto a RAM disk or
better into ODBC? All calls enter and maintain a record into the database
of their existence, perhaps that's your bottleneck.

Read
http://wiki.freeswitch.org/wiki/Performance_testing_and_configurations#Configurations
for hints



On 14 August 2014 19:59, Andre <andretodd at verizon.net> wrote:

> Just testing for now, no production traffic. Trying to find the breaking
> point.
>
>
>
> The reason for the questions are because I’m testing FS and after a
> certain point I am getting very high PDD, more than 6 seconds.  Per the
> CDR’s it takes about 6 seconds from start time in CDR Inbound to CDR
> outbound.
>
> I have tried 1 to 10 profiles to see if that makes a difference but it
> doesn’t and 1 to 4 consoles running one profile each still no difference.
>
>
>
> Any idea why the CDR’s should show it would take 6 seconds to get from
> start of inbound to start of outbound?
>
>
>
> *From:* freeswitch-users-bounces at lists.freeswitch.org [mailto:
> freeswitch-users-bounces at lists.freeswitch.org] *On Behalf Of *Ken Rice
> *Sent:* Thursday, August 14, 2014 2:06 PM
>
> *To:* FreeSWITCH Users Help
> *Subject:* Re: [Freeswitch-users] Network Card
>
>
>
> 1000 CPS the NIC is not the bottle neck this is actually relatively small
> amount of TCP/UDP traffic for just the SIP message.
>
> Now the RTP @ 1000 calls on the other hand assuming G711 at 20ms ptime.
> And assuming 1 leg calls originating from a dialer would generate
> 80megabits of RTP (at the IP layer even more on the wire), and approx 50K
> PPS. Assuming 2 leg’d calls bridging media 160megabit and 100K PPS of just
> RTP...
>
> But then we have to extrapolate peak concurrent from 1000 CPS... Assuming
> dialer traffic again, 75% ASR and 20 second ACD, and lets assume the 25%
> are just hard failed for the calculation (no early media on these) that
> gives us 750 new calls/sec and over the course of 20 seconds we accumulate
> 15,000 concurrent calls..
>
> So lets take that and extrapolate bandwidth and PPS requirements for
> single leg’d calls of 1.43Gig/sec (on the wire) and 750K PPS so that’s
> going to exceed the bandwidth of a GIG-E connection so you’d have to break
> this into 2 RTP endpoints in sofia, and make sure you switches/routers
> could handle the PPS loads.
>
> Double that to 2 leg calls and you have 2.86G/sec of RTP (again on the
> wire including UDP, IP, ethernet framing over heads) and pushing 1.5million
> PPS.
>
> Here again this dictates at least 3 gig-e NICs on your FS box. Is this
> possible? I’m sure it is with the right hardware configuration... But lets
> step back a minute and really consider what is going on...
>
> You are now putting approx over $2000/hr of traffic on a very expensive
> box that can and will at some point fail. Why are you doing such a thing?
> You’re better off limiting rates to sub 1000 CPS, (say 500CPS) deploying 3
> boxes to handle the load in n+1 configuration, and if you lose a box you
> arent losing $2000+ in revenue and cranky customers while you
> repair/replace the now dead machine.
>
> Lets face it, you can purchase a colo a nice machine for say 200 to
> 300$/month (assuming bandwidth costs are the same as above so that’s a wash
> instead of trying to cram it all on box that ends up costing you 1000/month
> by the time you pay for the hardware and colocation expenses for it.
>
>
>
>
>
> On 8/14/14 11:56 AM, "Russell Treleaven" <rtreleaven at bunnykick.ca> wrote:
>
> 1000 calls per second... no idea but I really doubt the network interface
> is going to be the bottleneck.
> Not sure why you make a distinction here between sessions  and connected.
> To be clear you mean 5000 bridge calls correct?
> Since media is not an issue I think you can just start testing with sipp
> and bwm-ng to see what is what.
>
> 1000 calls per second is a very big number.
> Let us know how you make out.
>
> Sincerely,
>
> Russell Treleaven
>
>
>
>
>
> On Wed, Aug 13, 2014 at 5:40 PM, Andre <andretodd at verizon.net> wrote:
>
> 1000 calls per second and 10000 sessions with 4000 connected. What would
> be recommend?
>
> *From:* freeswitch-users-bounces at lists.freeswitch.org [
> mailto:freeswitch-users-bounces at lists.freeswitch.org
> <freeswitch-users-bounces at lists.freeswitch.org>] *On Behalf Of *Russell
> Treleaven
> *Sent:* Wednesday, August 13, 2014 5:25 PM
> *To:* FreeSWITCH Users Help
> *Subject:* Re: [Freeswitch-users] Network Card
>
>
>
> If the media is bypassing you then you only need bandwidth for the
> signalling.
>
>
>
> Depends primarily on the calls per second.
>
>
>
> bwm-ng will help you to see whats going on.
>
>
>
>
>
>
>
>
>
> On Wed, Aug 13, 2014 at 5:20 PM, Andre <andretodd at verizon.net> wrote:
>
>
>
> Hi, is there anything special with the network card to get 10,000 sessions
> in media bypass mode with FreeSwitch?  Will a Gig E be enough bandwidth?
>
> What would it look like if the network card couldn’t keep up and how can I
> verify the network card is the bottleneck?
> Thanks
> Andre
>
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> 
> 
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://wiki.freeswitch.org
> http://www.cluecon.com
>
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
> http://www.freeswitch.org
>
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> 
> 
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://wiki.freeswitch.org
> http://www.cluecon.com
>
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
> http://www.freeswitch.org
>
>
> ------------------------------
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> 
> 
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://wiki.freeswitch.org
> http://www.cluecon.com
>
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
> http://www.freeswitch.org
>
>
> --
> Ken
>
>
>
> *http://www.FreeSWITCH.org
> <http://www.FreeSWITCH.org>http://www.ClueCon.com <http://www.ClueCon.com>
> http://www.OSTAG.org <http://www.OSTAG.org>*irc.freenode.net #freeswitch
> Twitter: @FreeSWITCH
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> 
> 
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://wiki.freeswitch.org
> http://www.cluecon.com
>
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
> http://www.freeswitch.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20140815/fae72f20/attachment.html 


Join us at ClueCon 2016 Aug 8-12, 2016
More information about the FreeSWITCH-users mailing list