On Aug 1, 2009, at 12:02 PM, Alex Balashov wrote:
Scott Berkman wrote:
Asterisk also makes for a very powerful and scriptable IVR platform if you are willing/able to put the work in.
I was about to say - it makes an excellent feature box. My only major concern relates to its scalability; not even so much the amount of concurrent calls per box that it can handle as how much that amount might be reduced by a weighty, complicated dial plan with database backing, etc.
Scalability is the biggest issue. I'm lucky to get 200 - 250 calls per box with some pretty heave DB and playing some sound files/IVR prompts. But being able to put it on commodity hardware makes this pretty easy to deal with if you are accustomed/equipped to mass manage/ monitor linux boxen. I've got 15 of them online now.....and adding more is just a matter of provisioning one of the spare blades. As far as upgrades go, we pretty well steer clear. It seems that the realtime architecture and other DB functions are not at the top of the dev's lists, nor do they even have the capability of testing these things under load. Most upgrades have been massive amounts of pain until we learned the easy trick: dedicate one production box to whatever new version seems to pass muster in staging. Dump traffic on it. If it doesn't work, don't sweat. Start over with staging again in a couple of week - it's alway like a brand new product. The bugs you had are probably gone, and they are somewhere else that you don't care about now. You pay in licensing, or you pay in time. It's your call. For now, we're paying in time. Daryl