[asterisk-users] can we install 10 PCI card on asterisk

[Shaun Ruffell]

On Mon, Aug 27, 2012 at 06:10:23PM +0530, Raj Mathur (राज माथुर) wrote:
> On Monday 27 Aug 2012, DHAVAL INDRODIYA wrote:
> > i would like to know if anyone has done or having idea regarding PRI
> > terminations in asterisk.
> > 
> > i have a requirement where i need to support 80 PRI in one machine i
> > have found a machine which have 10 PCI slots available
> > 
> > now i am thinking of arranging 8port sangoma card in this pci slots
> > so i can arrenge 10 card in that.
> 
> Last I checked, the highest channel number DAHDI supported was 1023, 
> limiting you to some 34 E1 PRIs.

Since DAHDI-Linux 2.5.0 there is not a hard coded limit on the
number of channels [1] or spans [2] that DAHDI can support. The
number of pseudo channels is limited by a module parameter,
'max_pseudo_channels', which currently defaults to 512 [3].

[1] http://svnview.digium.com/svn/dahdi?view=revision&revision=9609
[2] http://svnview.digium.com/svn/dahdi?view=revision&revision=9598
[3] http://svnview.digium.com/svn/dahdi?view=revision&revision=9610

> > is it possible to run system like that ? is it good idea , can
> > asterisk handle 2400 calls if machine size and RAM is good.
> 
> We've faced stability issues with more than 500 simultaneous calls on a 
> single high-powered server, with no transcoding.  However, that's 
> probably more a limitation of our own architecture and application than 
> a hard Asterisk limit.

I too would be very surprised if DHAVAL is able to run anything
close to 80 spans in a single server without modifying the drivers
to a) optimize for increased throughput versus minimal latency
(DAHDI currently favors low latency between bridged channels) and b)
Rework how conferencing / bridging is done in the drivers so that
only channels in conference with one another are checked for audio
mixing. I believe the limiting factor would be CPU cycles and not
memory.

As an aside, the most I've run in a single server are 24 T1s. But I
was not optimizing for density either.

Cheers,
Shaun

-- 
Shaun Ruffell
Digium, Inc. | Linux Kernel Developer
445 Jan Davis Drive NW - Huntsville, AL 35806 - USA
Check us out at: www.digium.com & www.asterisk.org