[Asterisk-Users] 5,000 concurrent calls system rollout question

Chris A. Icide chris at netgeeks.net
Wed Feb 1 01:39:19 MST 2006


Even if you could, you wouldn't want to use just one system to handle
this call load.  What happens when you lose a power supply or a hard
drive, or any other random failure?

I would think you would want a more robust design.  While you can go the
signate way and use SGI hardware to increase your load per footprint,
you can alos go the way of a large cluster of low priced systems as well.

I would do something like this:

Two SIP router systems (all signalling, no media) that all SIP devices
(end user UA's provider trunks etc.) communicate with in a load balanced
fashion.  These two routers recieve registrations all SIP signalling. 
They keep track of dynamic UA locations (SER or Asterisk could be used
here).  They use a SIP 302 redirect where possible and re-invite where
redirect isn't supported to route call requests to a cluster of asterisk
systems.  For 5000 calls with no media, two systems should be good
enough for N+1 redundancy (in other words one server is enough, but you
have two so you can fail one at any time).

Behind this you stick as many asterisk servers as is needed based upon
the hardware and it's load ability.  Again, N+1 should be your minimum
design basis for the number of systems.  The two routing systems should
have a method of knowing the load on each node so that when redirecting
a call, they can do so intelligently.  This would also allow you to
build in the ability to take nodes offline for maintenance or other
requirements. 

Just throwing together a bunch of asterisk systems and using
'round-robin' routing will quickly become a management nightmare.

While this can definately be done using asterisk, like someone else
said, if you want to do it right, you are going to be looking at the
need for a strong implementation team.

-Chris




More information about the asterisk-users mailing list