[asterisk-users] 2000+ user Asterisk PBX

[Grey Man]

On Sat, Aug 2, 2008 at 2:53 PM, Femi <adalemo at gmail.com> wrote:
> Any 2000+ user Asterisk PBX installs out there?
>
> Please hit me off-list, I need some support on a 2000+ user Asterisk PBX
> with high availability and over 10E1s to PTOs
>

If you're talking about 2000+ SIP users then I have some experience.
When we got to 1000 users we replaced the Digium E1 cards with a Cisco
AS5400 to make our echo and static problems go away (that was 3 years
ago so the cards may have improved since then). Now we've largely
replaced the Cisco AS5400 with an SS7 switch but that was more a
business call related to supplier interconnect considerations than a
technical one. Those Cisco AS5400's are fantastic pieces of kit, it's
the only thing in our set up that we have never had even the smallest
problem with in over 3 years! An AS5400 will be more expensive than a
commodity server with 2 or 3 four port E1 cards which is the main
drawback. At the time it was more expensive for us to be continuously
troubleshooting static and echo issues and as mentioned above that
situation may have improved.

If you are talking SIP users the other thing you'll have to do is to
move as much of the non-call related SIP traffic off Asterisk. That
means a separate SIP Proxy and/or Registrar. We have split those two
functions off onto separate boxes so that Asterisk only has to deal
with SIP for call signalling and of course media which is what its big
strength is.

Finally you will need a nice big database box, realtime is the only
practical way to run Asterisk for anything over a few hundred users.
The SIP Registrar and Asterisk will both generate large loads on your
database and it is the critical link in the chain. If your database
has problems you can't get any calls out and if that's not bad enough
all your user's ATAs registrations will drop off meaning you get a
deluge of support calls. We use Postgresql which does a good job but
the big problem with it is redundancy. Postgresql does not really have
an industrial strength replication solution which means the time it
takes to switch over from the primary to secondary database is a
problem. MySQL seems to have a much better replication solution and
Asterisk doesn't really need a lot of the advantages Postgresql has
over MySQL such as better stored procedure support etc.

So in summary the critical factors in my opinion are:

1. Good database,
2. Good quality solution ofr E1's,
3. Split off non-media related signalling from Asterisk.

If you're not talking about 2000+ SIP users then you can pretty much
disregard everything I've said :-).

Regards,

Greyman.