[Asterisk-Users] asterisk with big number of extentions.

[Rich Adamson]

> > > Can Asterisk handle such kind of load?
> >
> > No problem, as long as none of them make any calls. What you really wanted
> > to ask is, how many simultanous calls can a single * box handle, and that
> > is a function of exactly "how" you configure/engineer it:
> >   1) sip phones that have canreinvite=yes use * for call setup only
> >   2) sip phones that have canreinvite=no force the rtp traffic through *
> >      consuming a greater amount of processing resources
> >   3) sip phones that are configured with different encoding (codec) schemes
> >      are forced through some translations somewhere. Is that * in your
> > case? If so, see #2.
> >   4) some sip phones (or adapters) don't handle sip invites very well, and
> >      must pass all rtp traffic through *. If you use those, see #2.
> >   5) some sip phones lurking behind NAT boxes require all rtp traffic to
> >      pass through *. See #2.
> >
> 
> Rick, thanks for reply. It looks like we are comming to #2 with our 
> configuration. At least half of the traffic will pass through our * box.
> see below...
> 
> > > Can it be done on one server?
> >
> > Probably not given the answer above.
> >
> > > What kind of server do we need for making it work?
> >
> > High reliability, fast processor, disk speed less important, high
> > throughput ethernet, and a good consultant or two to help. ;) See the wiki
> > pages.
> >
> 
> I assume you mean voip-info.org.  I've check hardware recomendations there. 
> But listed examples looks far away from the setup we are thinking about. The 
> biggest is about hundreds of phones.
> 
> It'll be interesting to get some real numbers how much simultaneous 
> connections (channels) certain server can handle.
> 
> Also I'm interesting in the ways how asterisk setup can be scaled for 
> extending number of channels.
> 
> > To properly engineer "any" pbx, you will have to start with some reasonable
> > call/traffic statistics, and engineer around those numbers. Asterisk is no
> > different.
> >
> 
> Thanks for advise. Unfortunately we don't have this kind of data for this 
> project. All we have is estimations.

Dmitry,
 
Estimations are fine for starters. Got to start somewhere, but that somewhere
has to be in terms of number of calls (not extensions) and other traffic/usage
statistics.

Someone on the list was messing around with load testing asterisk. You should
be able to find some reference in the list archives. I don't think I'd trust
the numbers though unless those load tests happen to use the same mix of
sip phones, translations, etc. If you really need to address 25,000 extensions,
then I'd be doing some rather serious well-thought-out hands-on load testing
to better understand how some of these config options impact performance, etc.
Think you might bump into a surprise or two that just may impact other parts
of your project plan.

Rich