[asterisk-users] Asterisk concurrent calls count

[Steve Totaro]

On Sat, May 17, 2008 at 3:11 PM, Mike Trest - On Travel <Mike at trest.com> wrote:
> At 11:44 AM 5/16/2008, you wrote:
>>Yes, you could probably add 2 or 3 or 10 or 15 to the number of calls
>>that a particular machine could handle, but from a support perspective, it
>>doesn't matter how many the machine could theoretically handle, it matters
>>how many it could handle in the particular installation in a supportable
>>configuration (those are all those pesky variables we've been talking about).
>
> Absolutely!   Right On!   Tell it like it is!    And many other
> cryptic encouragements.
>
> With very large scale deployments, I have a set of numbers available
> in my head
> that work well to predict how many machines will be needed for a
> particular application
> but I wind up being surprised by non-predictable "rate of arrival"  issues.
>
> Since most of my deployments are tied with Television and other promotional
> support, a single reference by the on-screen (or on-radio) commentator, and the
> phones are instantly flooded with thousands of new call setup
> requests.  Indeed,
> one such incident in a NASCAR race with 13M viewers, produced 18,000 new calls
> within two minutes.   The rate of arrival of new calls was dispersed
> to a farm of 60
> Asterisk in three widely separated regions of the
> US.   However,  approximately
> 15,000 calls were actually dropped on the PSTN / SS7 network before
> ever reaching
> three dispersed Asterisk farms.
>
> Those farms were being "fed" inbound calls by a network of
> 250+  Nortel switches with
> millions of subscribers.   However, the Los Angeles area PSTN network
> access facility
> had only 900 spare channels available in that two minute
> period.   Meanwhile, every asterisk
> answered every call and joint the callers into appropriate conference
> groups until every
> single available port was fully occupied.     This illustrates that
> such issues of call capacity
> exist completely apart from the Asterisk or whatever machine is used
> for implementation.
>
> So everyone should not be surprised by "it depends" kinds of answers
> to the question
> of concurrent call counts.  This application was so far off the
> typical product specifications
> that nothing published by Digium or anyone else could anticipate
> those surprises that
> come when you least expect.
>
> ..mike..
>

I don't think anyone is expecting any rough numbers from Digium about
the telco's ingress/egress.

Thanks,
Steve Totaro