[Asterisk-Users] Large Enterprises using asterisk

[Steven Critchfield]

On Thu, 2004-07-22 at 20:39, Tenorio, Leandro wrote:
> 	What I also saw in my little research that * in not suitable for
> large deployments like medium or large enterprises or sometimes even
> smalls ones with specifics needs, any of you could mention a lot of
> them, Call Centers, Banks, etc. and why not carriers for their own use.

<snipped to the appearant meat>

> 	IMHO, * need a change in the architecture, in the way it's made,
> and in the way it works, not just more features or Bug solving.
> 	I'll be thinking, yes I also think, in the last weeks the way to
> help, I'm not a programmer so I could not develop, I don't even know
> Linux the way you do, so I could not help that way either, but I could
> help in other ways, That's why I take 2hs to write this email and I love
> to do it.
> 	Think on it. Think on an * system with servers for Trunks (IP,
> TDM, don't care), servers for core routing, servers for VM, servers used
> as IVRs, remember what those companies needs (redundancy, fault
> tolerance, load balancing, among other things) * don't have these
> features yet, think on systems with 600, 1K or even 3K users, think on
> systems that when a server crash (and believe, servers always crash)
> others take the place, with a little o even better with no downtime, 4
> those users, trunks or systems. When that time comes, and trust me it
> could be done, * could be used in orgs. 

Maybe you still have more to read and learn. Luckily there are a lot of
things here to help. I have to spend time thinking about fail over
support due to the business I am in. I also have to think about
scalability as any growth requires us to scale up to larger hardware.

Since there are fail over switches for T1(links on wiki), your trunks
can be shuffled off one machine and onto another without too much
hassle. Next decidion is how much "failure" are you willing to deal with
in a single moment. It would be possible to set up 4 T1s into one
machine and all fail to a backup machine. It is possible to split the
T1s between tha machines with each being a backup for the other. With 4
machines you could have 1 line into each and a failover onto a
neighboring machine. 

As you can see, you can plan for a server failure. With IP extensions
that support more than one gateway, you can even have them work in a
fail over manner. You can even have the gateways be seperate from your
T1 gateways. If you support reinvites or transfers, you can have your
gateway just link phones to phones or phones to PSTN trunks. If the
gateway goes down, it can have some heartbeat software to let you swap a
new machine in automatically.

So how is it that asterisk needs to be reachitected? The peoblems you
describe are just a question of how to deploy, not how asterisk works.

If you deploy with fail over support, you should be able to achieve 4 or
5 9's and be on par with any other telecom hardware. Just know that it
will cost money.  
-- 
Steven Critchfield <critch at basesys.com>