[asterisk-users] Is Asterisk ready for Prime-Time?

[Robert Lister]

On Wed, Mar 19, 2008 at 11:43:21AM -0500, Bill Andersen wrote:
> This is not a troll.  I've used my real email because I want this
> taken seriously.  I'm not trying to make anyone mad, I just want
> some real discussion on this issue.  Please bare with me...
> 
> 2) Are there any users out there that really DO have an Asterisk
>    system that just works like clockwork?  I'm saying, once setup,
>    run for a year (or more) without any issues?


> 3) If SO, Should I simply consider a different vendor?

It depends. As they say, "Your Mileage May Vary"

You have gone with a pre-built asterisk based solution rather than rolling 
your own with 'plain' asterisk system. So without knowing your particular 
environment, it's obviously difficult to comment.

By the sound of it, your experience of asterisk has been based on one 
particular integrator's build of it.

One or two versions of asterisk out there were lemons and were best avoided.

And then there are some modules which are less stable than others. I have 
found that most of the core asterisk stuff to be reasonably stable and well 
behaved, but there are a few modules that either have problems, or have had 
problems in the past, which have now been fixed. chan_agent was a good 
example of something that worked on a small scale but certain bits of it 
were just broken.

Other problems may be down to operating system, memory, hardware 
or driver issues.

Here, I am using exclusively SIP devices, SIP media gateways (rather than PC 
hardware) with asterisk voicemail module and seems pretty stable. (We had to 
reboot the box 9 weeks ago for a kernel security update.)

pink*CLI> show uptime
System uptime: 9 weeks, 4 days, 23 hours, 44 minutes, 22 seconds

We have about 77 SIP devices and these are a mixture of hard 
and soft phones, with four media gateways. Spread over 9 sites.

There are a few ongoing intermittent issues, but haven't had any 
spontaneous crashes so far.


> 4) If NOT, and if my expectations are that a system SHOULD just
>    run and run without any problems.  Is Asterisk simply not my
>    solution.  Is Asterisk not REALLY ready for production.  Because
>    in my mind (as a user of phone services), "dealing" with the
>    phone system, even on a MONTHLY basis, means that the system

We did evaluate a number of other systems before we decided to go down the 
route of just plain asterisk and rolling our own, as nothing quite did what 
we wanted.

You could look at OpenSER but I'm not convinced you'd find that an easy 
thing to work with, when you describe what you want to achieve.

SipX was also pretty good, but these are SIP only servers rather than 
asterisk's multi-protocol ability (You also have to provide SIP media 
gateways rather than talk directly to a card in the back of the machine)

http://www.sipfoundry.org/sipX

SIPx is the open source release of Pingtel's SIPEchange product, which I 
also evaluated. it seemed like a pretty good 'set and forget' solution, and 
they are also now selling an integrated SIPx appliance:
http://www.patton.com/products/pe_products.asp?category=348&tab=fb&

Which we looked at and was pretty good. Up to 30 users and included 
automatic handset provisioning, nice GUI for setting things up etc. This is 
great where you have an environment where running a server is not possible. 

(our asterisk server is hosted in a nice data air conditioned centre with 
redundant disks, power, UPS, network.. everything, but no everybody can run 
an environment for ultra reliable servers, so an "Asterisk Appliance" might 
be a way forward and requires no server housing capability and very little 
knowledge of the operating system etc.

It is very difficult to stop thinking 'old PBX', and start thinking "What is 
it we're trying to achieve?" If what you want is a PBX, go and buy one. It 
was a tricky journey from the old PBX system to asterisk VoIP, as there were 
certain expectations of the old system, and maintaining lots of 
functionality with the new handsets/asterisk.

The system that replaced our PBX doesn't have anything like as many call 
features as the old PBX did, but then again, most of these features were 
almost never used. But what we did gain was much more flexibility, choice of 
handsets/clients, connection to various VoIP networks, the possibility of 
remote workers, redundancy in the new system, and integration possibilities 
with existing systems that were completely impossible on the old PBX system. 
(Or were only possible for lots of money!)

Handsets are finally evolving now, trying to put in features that were 
present on old PBXs with 'traditional' paradigms like key and lamps etc, 
which users want on VoIP systems, but I believe that will ultimately lead to 
more proprietary systems and will ultimately fail in favour of Soft Phones, 
which are much better able to add new features rather than be constrained by 
a physical handset with buttons and memory limitations etc.

In my experience, you can buy a very expensive handset or a reasonably 
priced soft phone that can do exactly the same features. People are 
reluctant to give up their physical handsets on desks just yet though.

There is something to be said about keeping things simple. We decided not to 
go for the option of putting, say, BRI cards in the back of asterisk server 
and using it as a media gateway, but use separate off-the-shelf media 
gateway boxes to do this instead, so from the asterisk server, everything is 
SIP, and the media gateways are doing the awkward ISDN/Analogue <-> IP 
conversion. That way, any SIP server (including Asterisk) can interact with 
these gateways. Reading this list, some of the PCI cards and drivers are 
better/more stable than others.

The benefit you get with Asterisk/VoIP is that you are not locked in to one 
particular vendor, architecture or approach, and can be very cost effective 
for the equivalent commercial products you would have to buy to get the same 
functionality in the traditional PBX world. 

The trade-off having an open system and flexibility, is that there are so 
many possible configurations/handsets etc. that it can be a challenge to get 
it all working. If you want somebody to have done all the work for you and 
not to have to worry about it, then maybe go for a packaged, proprietary 
solution.

If lock-in is not an issue for you, then you could just buy a lucent PBX 
system or Cisco call manager, for example.

Another approach might be to go for a managed VoIP service provider, and 
just connect your handsets to their service rather than run your own 
servers. this way, you get the benefit of VoIP without having to run your 
own VoIP service.